The open-source data labeling tool market is experiencing robust growth, driven by the increasing demand for high-quality training data in various AI applications. The market's expansion is fueled by several key factors: the rising adoption of machine learning and deep learning algorithms across industries, the need for efficient and cost-effective data annotation solutions, and a growing preference for customizable and flexible tools that can adapt to diverse data types and project requirements. While proprietary solutions exist, the open-source ecosystem offers advantages including community support, transparency, cost-effectiveness, and the ability to tailor tools to specific needs, fostering innovation and accessibility. The market is segmented by tool type (image, text, video, audio), deployment model (cloud, on-premise), and industry (automotive, healthcare, finance). We project a market size of approximately $500 million in 2025, with a compound annual growth rate (CAGR) of 25% from 2025 to 2033, reaching approximately $2.7 billion by 2033. This growth is tempered by challenges such as the complexities associated with data security, the need for skilled personnel to manage and use these tools effectively, and the inherent limitations of certain open-source solutions compared to their commercial counterparts. Despite these restraints, the open-source model's inherent flexibility and cost advantages will continue to attract a significant user base. The market's competitive landscape includes established players like Alecion and Appen, alongside numerous smaller companies and open-source communities actively contributing to the development and improvement of these tools. Geographical expansion is expected across North America, Europe, and Asia-Pacific, with the latter projected to witness significant growth due to the increasing adoption of AI and machine learning in developing economies. Future market trends point towards increased integration of automated labeling techniques within open-source tools, enhanced collaborative features to improve efficiency, and further specialization to cater to specific data types and industry-specific requirements. Continuous innovation and community contributions will remain crucial drivers of growth in this dynamic market segment.

Image Annotation Tool Market Outlook

The global image annotation tool market size is projected to grow from approximately $700 million in 2023 to an estimated $2.5 billion by 2032, exhibiting a remarkable compound annual growth rate (CAGR) of 15.2% over the forecast period. The surging demand for machine learning and artificial intelligence applications is driving this robust market expansion. Image annotation tools are crucial for training AI models to recognize and interpret images, a necessity across diverse industries.

One of the key growth factors fueling the image annotation tool market is the rapid adoption of AI and machine learning technologies across various sectors. Organizations in healthcare, automotive, retail, and many other industries are increasingly leveraging AI to enhance operational efficiency, improve customer experiences, and drive innovation. Accurate image annotation is essential for developing sophisticated AI models, thereby boosting the demand for these tools. Additionally, the proliferation of big data analytics and the growing necessity to manage large volumes of unstructured data have amplified the need for efficient image annotation solutions.

Another significant driver is the increasing use of autonomous systems and applications. In the automotive industry, for instance, the development of autonomous vehicles relies heavily on annotated images to train algorithms for object detection, lane discipline, and navigation. Similarly, in the healthcare sector, annotated medical images are indispensable for developing diagnostic tools and treatment planning systems powered by AI. This widespread application of image annotation tools in the development of autonomous systems is a critical factor propelling market growth.

The rise of e-commerce and the digital retail landscape has also spurred demand for image annotation tools. Retailers are using these tools to optimize visual search features, personalize shopping experiences, and enhance inventory management through automated recognition of products and categories. Furthermore, advancements in computer vision technology have expanded the capabilities of image annotation tools, making them more accurate and efficient, which in turn encourages their adoption across various industries.

Data Annotation Software plays a pivotal role in the image annotation tool market by providing the necessary infrastructure for labeling and categorizing images efficiently. These software solutions are designed to handle various annotation tasks, from simple bounding boxes to complex semantic segmentation, enabling organizations to generate high-quality training datasets for AI models. The continuous advancements in data annotation software, including the integration of machine learning algorithms for automated labeling, have significantly enhanced the accuracy and speed of the annotation process. As the demand for AI-driven applications grows, the reliance on robust data annotation software becomes increasingly critical, supporting the development of sophisticated models across industries.

Regionally, North America holds the largest share of the image annotation tool market, driven by significant investments in AI and machine learning technologies and the presence of leading technology companies. Europe follows, with strong growth supported by government initiatives promoting AI research and development. The Asia Pacific region presents substantial growth opportunities due to the rapid digital transformation in emerging economies and increasing investments in technology infrastructure. Latin America and the Middle East & Africa are also expected to witness steady growth, albeit at a slower pace, due to the gradual adoption of advanced technologies.

Component Analysis

The image annotation tool market by component is segmented into software and services. The software segment dominates the market, encompassing a variety of tools designed for different annotation tasks, from simple image labeling to complex polygonal, semantic, or instance segmentation. The continuous evolution of software platforms, integrating advanced features such as automated annotation and machine learning algorithms, has significantly enhanced the accuracy and efficiency of image annotations. Furthermore, the availability of open-source annotation tools has lowered the entry barrier, allowing more organizations to adopt these technologies.

Services associated with image ann

The open-source data annotation tool market is experiencing robust growth, driven by the increasing demand for high-quality training data in artificial intelligence (AI) and machine learning (ML) applications. The market's expansion is fueled by several key factors: the rising adoption of AI across various industries (including automotive, healthcare, and finance), the need for efficient and cost-effective data annotation solutions, and a growing preference for flexible, customizable tools offered by open-source platforms. While cloud-based solutions currently dominate the market due to scalability and accessibility, on-premise deployments remain significant for organizations with stringent data security requirements. The competitive landscape is dynamic, with numerous established players and emerging startups vying for market share. The market is segmented geographically, with North America and Europe currently holding the largest shares due to early adoption of AI technologies and robust research & development activities. However, the Asia-Pacific region is projected to witness significant growth in the coming years, driven by increasing investments in AI infrastructure and talent development. Challenges remain, such as the need for skilled annotators and the ongoing evolution of annotation techniques to handle increasingly complex data types. The forecast period (2025-2033) suggests continued expansion, with a projected Compound Annual Growth Rate (CAGR) – let's conservatively estimate this at 15% based on typical growth in related software sectors. This growth will be influenced by advancements in automation and semi-automated annotation tools, as well as the emergence of novel annotation paradigms. The market is expected to see further consolidation, with larger players potentially acquiring smaller, specialized companies. The increasing focus on data privacy and security will necessitate the development of more robust and compliant open-source annotation tools. Specific application segments like healthcare, with its stringent regulatory landscape, and the automotive industry, with its reliance on autonomous driving technology, will continue to be major drivers of market growth. The increasing availability of open-source datasets and pre-trained models will indirectly contribute to the market’s expansion by lowering the barrier to entry for AI development.

Discover the booming market for open-source data labeling tools! Learn about its $500 million valuation in 2025, projected 25% CAGR, key drivers, and top players shaping this rapidly expanding sector within the AI revolution. Explore market trends and forecasts through 2033.

Leaves from genetically unique Juglans regia plants were scanned using X-ray micro-computed tomography (microCT) on the X-ray μCT beamline (8.3.2) at the Advanced Light Source (ALS) in Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA USA). Soil samples were collected in Fall of 2017 from the riparian oak forest located at the Russell Ranch Sustainable Agricultural Institute at the University of California Davis. The soil was sieved through a 2 mm mesh and was air dried before imaging. A single soil aggregate was scanned at 23 keV using the 10x objective lens with a pixel resolution of 650 nanometers on beamline 8.3.2 at the ALS. Additionally, a drought stressed almond flower bud (Prunus dulcis) from a plant housed at the University of California, Davis, was scanned using a 4x lens with a pixel resolution of 1.72 µm on beamline 8.3.2 at the ALS Raw tomographic image data was reconstructed using TomoPy. Reconstructions were converted to 8-bit tif or png format using ImageJ or the PIL package in Python before further processing. Images were annotated using Intel’s Computer Vision Annotation Tool (CVAT) and ImageJ. Both CVAT and ImageJ are free to use and open source. Leaf images were annotated in following Théroux-Rancourt et al. (2020). Specifically, Hand labeling was done directly in ImageJ by drawing around each tissue; with 5 images annotated per leaf. Care was taken to cover a range of anatomical variation to help improve the generalizability of the models to other leaves. All slices were labeled by Dr. Mina Momayyezi and Fiona Duong.To annotate the flower bud and soil aggregate, images were imported into CVAT. The exterior border of the bud (i.e. bud scales) and flower were annotated in CVAT and exported as masks. Similarly, the exterior of the soil aggregate and particulate organic matter identified by eye were annotated in CVAT and exported as masks. To annotate air spaces in both the bud and soil aggregate, images were imported into ImageJ. A gaussian blur was applied to the image to decrease noise and then the air space was segmented using thresholding. After applying the threshold, the selected air space region was converted to a binary image with white representing the air space and black representing everything else. This binary image was overlaid upon the original image and the air space within the flower bud and aggregate was selected using the “free hand” tool. Air space outside of the region of interest for both image sets was eliminated. The quality of the air space annotation was then visually inspected for accuracy against the underlying original image; incomplete annotations were corrected using the brush or pencil tool to paint missing air space white and incorrectly identified air space black. Once the annotation was satisfactorily corrected, the binary image of the air space was saved. Finally, the annotations of the bud and flower or aggregate and organic matter were opened in ImageJ and the associated air space mask was overlaid on top of them forming a three-layer mask suitable for training the fully convolutional network. All labeling of the soil aggregate and soil aggregate images was done by Dr. Devin Rippner. These images and annotations are for training deep learning models to identify different constituents in leaves, almond buds, and soil aggregates Limitations: For the walnut leaves, some tissues (stomata, etc.) are not labeled and only represent a small portion of a full leaf. Similarly, both the almond bud and the aggregate represent just one single sample of each. The bud tissues are only divided up into buds scales, flower, and air space. Many other tissues remain unlabeled. For the soil aggregate annotated labels are done by eye with no actual chemical information. Therefore particulate organic matter identification may be incorrect. Resources in this dataset:Resource Title: Annotated X-ray CT images and masks of a Forest Soil Aggregate. File Name: forest_soil_images_masks_for_testing_training.zipResource Description: This aggregate was collected from the riparian oak forest at the Russell Ranch Sustainable Agricultural Facility. The aggreagate was scanned using X-ray micro-computed tomography (microCT) on the X-ray μCT beamline (8.3.2) at the Advanced Light Source (ALS) in Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA USA) using the 10x objective lens with a pixel resolution of 650 nanometers. For masks, the background has a value of 0,0,0; pores spaces have a value of 250,250, 250; mineral solids have a value= 128,0,0; and particulate organic matter has a value of = 000,128,000. These files were used for training a model to segment the forest soil aggregate and for testing the accuracy, precision, recall, and f1 score of the model.Resource Title: Annotated X-ray CT images and masks of an Almond bud (P. Dulcis). File Name: Almond_bud_tube_D_P6_training_testing_images_and_masks.zipResource Description: Drought stressed almond flower bud (Prunis dulcis) from a plant housed at the University of California, Davis, was scanned by X-ray micro-computed tomography (microCT) on the X-ray μCT beamline (8.3.2) at the Advanced Light Source (ALS) in Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA USA) using the 4x lens with a pixel resolution of 1.72 µm using. For masks, the background has a value of 0,0,0; air spaces have a value of 255,255, 255; bud scales have a value= 128,0,0; and flower tissues have a value of = 000,128,000. These files were used for training a model to segment the almond bud and for testing the accuracy, precision, recall, and f1 score of the model.Resource Software Recommended: Fiji (ImageJ),url: https://imagej.net/software/fiji/downloads Resource Title: Annotated X-ray CT images and masks of Walnut leaves (J. Regia) . File Name: 6_leaf_training_testing_images_and_masks_for_paper.zipResource Description: Stems were collected from genetically unique J. regia accessions at the 117 USDA-ARS-NCGR in Wolfskill Experimental Orchard, Winters, California USA to use as scion, and were grafted by Sierra Gold Nursery onto a commonly used commercial rootstock, RX1 (J. microcarpa × J. regia). We used a common rootstock to eliminate any own-root effects and to simulate conditions for a commercial walnut orchard setting, where rootstocks are commonly used. The grafted saplings were repotted and transferred to the Armstrong lathe house facility at the University of California, Davis in June 2019, and kept under natural light and temperature. Leaves from each accession and treatment were scanned using X-ray micro-computed tomography (microCT) on the X-ray μCT beamline (8.3.2) at the Advanced Light Source (ALS) in Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA USA) using the 10x objective lens with a pixel resolution of 650 nanometers. For masks, the background has a value of 170,170,170; Epidermis value= 85,85,85; Mesophyll value= 0,0,0; Bundle Sheath Extension value= 152,152,152; Vein value= 220,220,220; Air value = 255,255,255.Resource Software Recommended: Fiji (ImageJ),url: https://imagej.net/software/fiji/downloads

The data annotation and labeling tools market is experiencing robust growth, driven by the escalating demand for high-quality training data in the burgeoning fields of artificial intelligence (AI) and machine learning (ML). The market's expansion is fueled by the increasing adoption of AI across diverse sectors, including autonomous vehicles, healthcare, and finance. These industries require vast amounts of accurately labeled data to train their AI models, leading to a significant surge in the demand for efficient and scalable annotation tools. While precise market sizing for 2025 is unavailable, considering a conservative estimate and assuming a CAGR of 25% (a reasonable figure given industry growth), we can project a market value exceeding $2 billion in 2025, rising significantly over the forecast period (2025-2033). Key trends include the growing adoption of cloud-based solutions, increased automation in the annotation process through AI-assisted tools, and a heightened focus on data privacy and security. The rise of synthetic data generation is also beginning to impact the market, offering potential cost savings and improved data diversity. However, challenges remain. The high cost of skilled annotators, the need for continuous quality control, and the inherent complexities of labeling diverse data types (images, text, audio, video) pose significant restraints on market growth. While leading players like Labelbox, Scale AI, and SuperAnnotate dominate the market with advanced features and robust scalability, smaller companies and open-source tools continue to compete, often focusing on niche applications or offering cost-effective alternatives. The competitive landscape is dynamic, with continuous innovation and mergers and acquisitions shaping the future of this rapidly evolving market. Regional variations in adoption are also expected, with North America and Europe likely leading the market, followed by Asia-Pacific and other regions. This continuous evolution necessitates careful strategic planning and adaptation for businesses operating in or considering entry into this space.

Automated Image Annotation for Microscopy Market Outlook

According to our latest research, the global Automated Image Annotation for Microscopy market size reached USD 542.7 million in 2024, reflecting robust adoption across life sciences and healthcare research. The market is projected to expand at a CAGR of 18.2% from 2025 to 2033, with the total market value anticipated to reach USD 2,464.8 million by 2033. This remarkable growth is being driven by the increasing demand for high-throughput, accurate, and scalable image analysis solutions in medical diagnostics, pharmaceutical research, and academic settings.

The primary growth factor propelling the Automated Image Annotation for Microscopy market is the exponential rise in the volume and complexity of microscopy image data generated in life sciences research and clinical diagnostics. As advanced imaging modalities such as confocal, super-resolution, and electron microscopy become commonplace, researchers face mounting challenges in manually annotating vast datasets. Automated image annotation platforms, leveraging artificial intelligence and deep learning, provide significant efficiency gains by streamlining annotation workflows, minimizing human error, and enabling reproducible data labeling at scale. This technological leap is particularly critical in fields like cell biology, pathology, and neuroscience, where precise annotation is essential for downstream analysis, disease modeling, and biomarker discovery.

Another key driver is the growing integration of automated annotation tools into end-to-end digital pathology and drug discovery pipelines. Pharmaceutical and biotechnology companies are increasingly investing in automation to accelerate preclinical research, reduce time-to-market for new therapeutics, and enhance the reliability of high-content screening assays. Automated image annotation not only expedites the identification and classification of cellular structures but also supports quantitative analysis required for regulatory submissions and clinical trials. Furthermore, the rising adoption of cloud-based platforms is democratizing access to advanced annotation tools, enabling collaboration across geographically dispersed research teams and facilitating the aggregation of large annotated datasets for AI model training.

The market is also benefitting from significant advancements in machine learning algorithms, including semantic segmentation, instance segmentation, and object detection, which have dramatically improved annotation accuracy and versatility. These innovations are reducing the barriers for adoption among academic and research institutions, which often operate under tight resource constraints. Additionally, the increasing prevalence of open-source annotation frameworks and interoperability standards is fostering an ecosystem where automated annotation solutions can be seamlessly integrated with existing microscopy workflows. As a result, the Automated Image Annotation for Microscopy market is poised for sustained growth, with emerging applications in personalized medicine, digital pathology, and precision oncology further expanding its addressable market.

From a regional perspective, North America currently leads the global Automated Image Annotation for Microscopy market, accounting for the largest share in 2024, followed closely by Europe and Asia Pacific. The dominance of North America is attributed to the high concentration of pharmaceutical companies, advanced healthcare infrastructure, and significant investments in AI-driven healthcare solutions. However, Asia Pacific is expected to witness the fastest growth during the forecast period, driven by increasing R&D expenditure, expanding biotechnology sectors, and rising adoption of digital pathology solutions in countries such as China, Japan, and India. This regional diversification is expected to fuel market expansion and foster innovation in automated image annotation technologies worldwide.

Component Analysis

The Automated Image Annotation for

DESCRIPTION
For this task, we use a subset of the MIRFLICKR (http://mirflickr.liacs.nl) collection. The entire collection contains 1 million images from the social photo sharing website Flickr and was formed by downloading up to a thousand photos per day that were deemed to be the most interesting according to Flickr. All photos in this collection were released by their users under a Creative Commons license, allowing them to be freely used for research purposes. Of the entire collection, 25 thousand images were manually annotated with a limited number of concepts and many of these annotations have been further refined and expanded over the lifetime of the ImageCLEF photo annotation task. This year we used crowd sourcing to annotate all of these 25 thousand images with the concepts.

On this page we provide you with more information about the textual features, visual features and concept features we supply with each image in the collection we use for this year's task.

TEXTUAL FEATURES
All images are accompanied by the following textual features:

- Flickr user tags
These are the tags that the users assigned to the photos their uploaded to Flickr. The 'raw' tags are the original tags, while the 'clean' tags are those collapsed to lowercase and condensed to removed spaces.

- EXIF metadata
If available, the EXIF metadata contains information about the camera that took the photo and the parameters used. The 'raw' exif is the original camera data, while the 'clean' exif reduces the verbosity.

- User information and Creative Commons license information
This contains information about the user that took the photo and the license associated with it.

VISUAL FEATURES
Over the previous years of the photo annotation task we noticed that often the same types of visual features are used by the participants, in particular features based on interest points and bag-of-words are popular. To assist you we have extracted several features for you that you may want to use, so you can focus on the concept detection instead. We additionally give you some pointers to easy to use toolkits that will help you extract other features or the same features but with different default settings.

- SIFT, C-SIFT, RGB-SIFT, OPPONENT-SIFT
We used the ISIS Color Descriptors (http://www.colordescriptors.com) toolkit to extract these descriptors. This package provides you with many different types of features based on interest points, mostly using SIFT. It furthermore assists you with building codebooks for bag-of-words. The toolkit is available for Windows, Linux and Mac OS X.

- SURF
We used the OpenSURF (http://www.chrisevansdev.com/computer-vision-opensurf.html) toolkit to extract this descriptor. The open source code is available in C++, C#, Java and many more languages.

- TOP-SURF
We used the TOP-SURF (http://press.liacs.nl/researchdownloads/topsurf) toolkit to extract this descriptor, which represents images with SURF-based bag-of-words. The website provides codebooks of several different sizes that were created using a combination of images from the MIR-FLICKR collection and from the internet. The toolkit also offers the ability to create custom codebooks from your own image collection. The code is open source, written in C++ and available for Windows, Linux and Mac OS X.

- GIST
We used the LabelMe (http://labelme.csail.mit.edu) toolkit to extract this descriptor. The MATLAB-based library offers a comprehensive set of tools for annotating images.

For the interest point-based features above we used a Fast Hessian-based technique to detect the interest points in each image. This detector is built into the OpenSURF library. In comparison with the Hessian-Laplace technique built into the ColorDescriptors toolkit it detects fewer points, resulting in a considerably reduced memory footprint. We therefore also provide you with the interest point locations in each image that the Fast Hessian-based technique detected, so when you would like to recalculate some features you can use them as a starting point for the extraction. The ColorDescriptors toolkit for instance accepts these locations as a separate parameter. Please go to http://www.imageclef.org/2012/photo-flickr/descriptors for more information on the file format of the visual features and how you can extract them yourself if you want to change the default settings.

CONCEPT FEATURES
We have solicited the help of workers on the Amazon Mechanical Turk platform to perform the concept annotation for us. To ensure a high standard of annotation we used the CrowdFlower platform that acts as a quality control layer by removing the judgments of workers that fail to annotate properly. We reused several concepts of last year's task and for most of these we annotated the remaining photos of the MIRFLICKR-25K collection that had not yet been used before in the previous task; for some concepts we reannotated all 25,000 images to boost their quality. For the new concepts we naturally had to annotate all of the images.

- Concepts
For each concept we indicate in which images it is present. The 'raw' concepts contain the judgments of all annotators for each image, where a '1' means an annotator indicated the concept was present whereas a '0' means the concept was not present, while the 'clean' concepts only contain the images for which the majority of annotators indicated the concept was present. Some images in the raw data for which we reused last year's annotations only have one judgment for a concept, whereas the other images have between three and five judgments; the single judgment does not mean only one annotator looked at it, as it is the result of a majority vote amongst last year's annotators.

- Annotations
For each image we indicate which concepts are present, so this is the reverse version of the data above. The 'raw' annotations contain the average agreement of the annotators on the presence of each concept, while the 'clean' annotations only include those for which there was a majority agreement amongst the annotators.

You will notice that the annotations are not perfect. Especially when the concepts are more subjective or abstract, the annotators tend to disagree more with each other. The raw versions of the concept annotations should help you get an understanding of the exact judgments given by the annotators.

Automated Ultrastructure Annotation Software Market Outlook

According to our latest research, the global automated ultrastructure annotation software market size reached USD 1.42 billion in 2024, demonstrating a robust momentum driven by the increasing demand for advanced digital pathology and high-throughput life sciences research. The market is projected to expand at a CAGR of 12.7% from 2025 to 2033, reaching an estimated USD 4.19 billion by 2033. This impressive growth is primarily fueled by rapid advancements in artificial intelligence (AI), machine learning (ML), and imaging technologies, which are enabling more precise, efficient, and scalable annotation of ultrastructural data across medical and research domains.

One of the primary growth factors for the automated ultrastructure annotation software market is the surging adoption of AI-powered solutions in medical imaging and life sciences research. As the volume and complexity of ultrastructural data generated by high-resolution imaging techniques such as electron microscopy continue to escalate, traditional manual annotation methods are becoming increasingly unsustainable. Automated annotation software leverages sophisticated algorithms to deliver accurate, reproducible, and rapid analyses, significantly reducing human error and enhancing throughput. This capability is particularly critical in clinical diagnostics, drug discovery, and disease research, where timely and reliable data interpretation can directly impact patient outcomes and research productivity.

Another significant driver is the growing integration of automated ultrastructure annotation software into drug discovery pipelines and translational research. Pharmaceutical and biotechnology companies are increasingly relying on these advanced tools to accelerate the identification and characterization of cellular and subcellular structures, enabling more efficient target validation and compound screening. By automating the annotation process, organizations can substantially shorten research timelines, reduce operational costs, and improve the reproducibility of experimental results. Furthermore, regulatory agencies are endorsing digital pathology and automated image analysis, fostering a supportive environment for the widespread adoption of these solutions in both preclinical and clinical settings.

The rising prevalence of chronic diseases and the ongoing digital transformation of healthcare systems globally are also contributing to the market's expansion. Hospitals, clinics, and research institutions are investing heavily in state-of-the-art imaging infrastructure and informatics platforms to enhance diagnostic accuracy and support personalized medicine initiatives. The integration of automated ultrastructure annotation software with electronic health records (EHRs) and laboratory information management systems (LIMS) is streamlining workflows, facilitating interdisciplinary collaboration, and supporting data-driven decision-making. As healthcare providers increasingly recognize the value of automated annotation in improving patient care and operational efficiency, the demand for these solutions is expected to surge in the coming years.

Regionally, North America continues to dominate the global automated ultrastructure annotation software market, accounting for the largest revenue share in 2024. This leadership can be attributed to the presence of leading technology vendors, well-established healthcare infrastructure, and significant investments in biomedical research. Europe and Asia Pacific are also witnessing substantial growth, driven by increasing research funding, expanding healthcare IT adoption, and rising awareness of the benefits of automated imaging analysis. The Asia Pacific region, in particular, is expected to exhibit the highest CAGR over the forecast period, supported by the rapid development of healthcare and research ecosystems in countries such as China, Japan, and India.

Component Analysis

The automated ultrastructure annotation software market is segmented by component into software and services. The software segment encompasses proprietary and open-source platforms designed to automate the annotation of ultrastructural images, leveraging AI, ML, and advanced image processing algorithms. These solutions are increasingly being adopted across medical, research, and industrial settings due to their ability to enhance accuracy, scalability, and efficiency. The software segment currently holds the largest share of the market, as org

We provide annotated fish imagery data for use in deep learning models (e.g., convolutional neural networks) for individual and species recognition. For individual recognition models, the dataset consists of annotated .json files of individual brook trout imagery collected at the Eastern Ecological Science Center's Experimental Stream Laboratory. For species recognition models, the dataset consists of annotated .json files for 7 freshwater fish species: lake trout, largemouth bass, smallmouth bass, brook trout, rainbow trout, walleye, and northern pike. Species imagery was compiled from Anglers Atlas and modified to remove human faces for privacy protection. We used open-source VGG image annotation software developed by Oxford University: https://www.robots.ox.ac.uk/~vgg/software/via/via-1.0.6.html.

ABSTRACT An original dataset of thermal videos and images that simulate illegal movements around the border and in protected areas and are designed for training machines and deep learning models. The videos are recorded in areas around the forest, at night, in different weather conditions – in the clear weather, in the rain, and in the fog, and with people in different body positions (upright, hunched) and movement speeds (regu- lar walking, running) at different ranges from the camera. In addition to using standard camera lenses, telephoto lenses were also used to test their impact on the quality of thermal images and person detection in different weather conditions and distance from the camera. The obtained dataset comprises 7412 manually labeled images extracted from video frames captured in the long-wave infrared (LWIR) a segment of the electromagnetic (EM) spectrum.

Instructions:

About 20 minutes of recorded material from the clear weather scenario, 13 minutes from the fog scenario, and about 15 minutes from rainy weather were processed. The longer videos were cut into sequences and from these sequences individual frames were extracted, resulting in 11,900 images for the clear weather, 4,905 images for the fog, and 7,030 images for the rainy weather scenarios.

A total of 6,111 frames were manual annotated so that could be used to train the supervised model for person detection. When selecting the frames, it was taken into account that the selected frames include different weather conditions so that in the set there were 2,663 frames shot in clear weather conditions, 1,135 frames of fog, and 2,313 frames of rain.

The annotations were made using the open-source Yolo BBox Annotation Tool that can simultaneously store annotations in the three most popular machine learning annotation formats YOLO, VOC, and MS COCO so all three annotation formats are available. The image annotation consists of a centroid position of the bounding box around each object of interest, size of the bounding box in terms of width and height, and corresponding class label (Human or Dog).

This dataset corresponds to a collection of images and/or image-derived data available from National Cancer Institute Imaging Data Commons (IDC) [1]. This dataset was converted into DICOM representation and ingested by the IDC team. You can explore and visualize the corresponding images using IDC Portal here: https://portal.imaging.datacommons.cancer.gov/explore/filters/?analysis_results_id=RMS-Mutation-Prediction-Expert-Annotations.. You can use the manifests included in this Zenodo record to download the content of the collection following the Download instructions below.

Collection description

This dataset contains 2 components:

Annotations of multiple regions of interest performed by an expert pathologist with eight years of experience for a subset of hematoxylin and eosin (H&E) stained images from the RMS-Mutation-Prediction image collection [1,2]. Annotations were generated manually, using the Aperio ImageScope tool, to delineate regions of alveolar rhabdomyosarcoma (ARMS), embryonal rhabdomyosarcoma (ERMS), stroma, and necrosis [3]. The resulting planar contour annotations were originally stored in ImageScope-specific XML format, and subsequently converted into Digital Imaging and Communications in Medicine (DICOM) Structured Report (SR) representation using the open source conversion tool [4].

AI-generated annotations stored as probabilistic segmentations. WARNING: After the release of v20, it was discovered that a mistake had been made during data conversion that affected the newly-released segmentations accompanying the "RMS-Mutation-Prediction" collection. Segmentations released in v20 for this collection have the segment labels for alveolar rhabdomyosarcoma (ARMS) and embryonal rhabdomyosarcoma (ERMS) switched in the metadata relative to the correct labels. Thus segment 3 in the released files is labelled in the metadata (the SegmentSequence) as ARMS but should correctly be interpreted as ERMS, and conversely segment 4 in the released files is labelled as ERMS but should be correctly interpreted as ARMS. We apologize for the mistake and any confusion that it has caused, and will be releasing a corrected version of the files in the next release as soon as possible.

Many pixels from the whole slide images annotated by this dataset are not contained inside any annotation contours and are considered to belong to the background class. Other pixels are contained inside only one annotation contour and are assigned to a single class. However, cases also exist in this dataset where annotation contours overlap. In these cases, the pixels contained in multiple contours could be assigned membership in multiple classes. One example is a necrotic tissue contour overlapping an internal subregion of an area designated by a larger ARMS or ERMS annotation. The ordering of annotations in this DICOM dataset preserves the order in the original XML generated using ImageScope. These annotations were converted, in sequence, into segmentation masks and used in the training of several machine learning models. Details on the training methods and model results are presented in [1]. In the case of overlapping contours, the order in which annotations are processed may affect the generated segmentation mask if prior contours are overwritten by later contours in the sequence. It is up to the application consuming this data to decide how to interpret tissues regions annotated with multiple classes. The annotations included in this dataset are available for visualization and exploration from the National Cancer Institute Imaging Data Commons (IDC) 5 as of data release v18. Direct link to open the collection in IDC Portal: https://portal.imaging.datacommons.cancer.gov/explore/filters/?analysis_results_id=RMS-Mutation-Prediction-Expert-Annotations.

Files included

A manifest file's name indicates the IDC data release in which a version of collection data was first introduced. For example, pan_cancer_nuclei_seg_dicom-collection_id-idc_v19-aws.s5cmd corresponds to the annotations for th eimages in the collection_id collection introduced in IDC data release v19. DICOM Binary segmentations were introduced in IDC v20. If there is a subsequent version of this Zenodo page, it will indicate when a subsequent version of the corresponding collection was introduced.

For each of the collections, the following manifest files are provided:

rms_mutation_prediction_expert_annotations-idc_v20-aws.s5cmd: manifest of files available for download from public IDC Amazon Web Services buckets

rms_mutation_prediction_expert_annotations-idc_v20-gcs.s5cmd: manifest of files available for download from public IDC Google Cloud Storage buckets

rms_mutation_prediction_expert_annotations-idc_v20-dcf.dcf: Gen3 manifest (for details see https://learn.canceridc.dev/data/organization-of-data/guids-and-uuids)

Note that manifest files that end in -aws.s5cmd reference files stored in Amazon Web Services (AWS) buckets, while -gcs.s5cmd reference files in Google Cloud Storage. The actual files are identical and are mirrored between AWS and GCP.

Download instructions

Each of the manifests include instructions in the header on how to download the included files.

To download the files using .s5cmd manifests:

install idc-index package: pip install --upgrade idc-index

download the files referenced by manifests included in this dataset by passing the .s5cmd manifest file: idc download manifest.s5cmd

To download the files using .dcf manifest, see manifest header.

Acknowledgments

Imaging Data Commons team has been funded in whole or in part with Federal funds from the National Cancer Institute, National Institutes of Health, under Task Order No. HHSN26110071 under Contract No. HHSN261201500003l.

If you use the files referenced in the attached manifests, we ask you to cite this dataset, as well as the publication describing the original dataset [2] and publication acknowledging IDC [5].

References

[1] D. Milewski et al., "Predicting molecular subtype and survival of rhabdomyosarcoma patients using deep learning of H&E images: A report from the Children's Oncology Group," Clin. Cancer Res., vol. 29, no. 2, pp. 364–378, Jan. 2023, doi: 10.1158/1078-0432.CCR-22-1663.

[2] Clunie, D., Khan, J., Milewski, D., Jung, H., Bowen, J., Lisle, C., Brown, T., Liu, Y., Collins, J., Linardic, C. M., Hawkins, D. S., Venkatramani, R., Clifford, W., Pot, D., Wagner, U., Farahani, K., Kim, E., & Fedorov, A. (2023). DICOM converted whole slide hematoxylin and eosin images of rhabdomyosarcoma from Children's Oncology Group trials [Data set]. Zenodo. https://doi.org/10.5281/zenodo.8225132

[3] Agaram NP. Evolving classification of rhabdomyosarcoma. Histopathology. 2022 Jan;80(1):98-108. doi: 10.1111/his.14449. PMID: 34958505; PMCID: PMC9425116,https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9425116/

[4] Chris Bridge. (2024). ImagingDataCommons/idc-sm-annotations-conversion: v1.0.0 (v1.0.0). Zenodo. https://doi.org/10.5281/zenodo.10632182

[5] Fedorov, A., Longabaugh, W. J. R., Pot, D., Clunie, D. A., Pieper, S. D., Gibbs, D. L., Bridge, C., Herrmann, M. D., Homeyer, A., Lewis, R., Aerts, H. J. W. L., Krishnaswamy, D., Thiriveedhi, V. K., Ciausu, C., Schacherer, D. P., Bontempi, D., Pihl, T., Wagner, U., Farahani, K., Kim, E. & Kikinis, R. National cancer institute imaging data commons: Toward transparency, reproducibility, and scalability in imaging artificial intelligence. Radiographics 43, (2023).

The "Day-Old Chicken Dataset" is a collection of images that were taken in a poultry farm. The images were taken from boxes containing "day-old chicks" and include both top-view and individual images of the chicks. The images were manually labeled using an open-source online deep-learning annotation tool called Roboflow. The final dataset is split into a training set (70%), validation set (10%) and a test set (20%) for evaluating the performance of object detection models. The dataset can be used to train and evaluate models for counting the number of "day-old chicks" in a box, which can be useful for farmers to keep track of their inventory.

Ground-based observations from fixed-mount cameras have the potential to fill an important role in environmental sensing, including direct measurement of water levels and qualitative observation of ecohydrological research sites. All of this is theoretically possible for anyone who can install a trail camera. Easy acquisition of ground-based imagery has resulted in millions of environmental images stored, some of which are public data, and many of which contain information that has yet to be used for scientific purposes. The goal of this project was to develop and document key image processing and machine learning workflows, primarily related to semi-automated image labeling, to increase the use and value of existing and emerging archives of imagery that is relevant to ecohydrological processes.

   This data package includes imagery, annotation files, water segmentation model and model performance plots, and model test results (overlay images and masks) for USGS Monitoring Site East Branch Brandywine Creek below Downingtown, PA. All imagery was acquired from the USGS Hydrologic Imagery Visualization and Information System (HIVIS; see https://apps.usgs.gov/hivis/camera/PA_East_Branch_Brandywine_Creek_below_Downingtown for this specific data set) and/or the National Imagery Management System (NIMS) API.

   Water segmentation models were created by tuning the open-source Segment Anything Model 2 (SAM2, https://github.com/facebookresearch/sam2) using images that were annotated by team members on this project. The models were trained on the "water" annotations, but annotation files may include additional labels, such as "snow", "sky", and "unknown". Image annotation was done in Computer Vision Annotation Tool (CVAT) and exported in COCO format (.json).

   All model training and testing was completed in GaugeCam Remote Image Manager Educational Artificial Intelligence (GRIME AI, https://gaugecam.org/) software (Version: Beta 16). Model performance plots were automatically generated during this process.

   This project was conducted in 2023-2025 by collaborators at the University of Nebraska-Lincoln, University of Nebraska at Kearney, and the U.S. Geological Survey.

   This material is based upon work supported by the U.S. Geological Survey under Grant/Cooperative Agreement No. G23AC00141-00. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the opinions or policies of the U.S. Geological Survey. Mention of trade names or commercial products does not constitute their endorsement by the U.S. Geological Survey. We gratefully acknowledge graduate student support from Daugherty Water for Food Global Institute at the University of Nebraska.

Introduction:

https://qiangli.de/imgs/flowchart2%20(1).png">

🌟 XimageNet-12 🌟

An Explainable Visual Benchmark Dataset for Robustness Evaluation. A Dataset for Image Background Exploration!

Blur Background, Segmented Background, AI-generated Background, Bias of Tools During Annotation, Color in Background, Random Background with Real Environment

+⭐ Follow Authors for project updates.

Website: XimageNet-12

Here, we trying to understand how image background effect the Computer Vision ML model, on topics such as Detection and Classification, based on baseline Li et.al work on ICLR 2022: Explainable AI: Object Recognition With Help From Background, we are now trying to enlarge the dataset, and analysis the following topics: Blur Background / Segmented Background / AI generated Background/ Bias of tools during annotation/ Color in Background / Dependent Factor in Background/ LatenSpace Distance of Foreground/ Random Background with Real Environment! Ultimately, we also define the math equation of Robustness Scores! So if you feel interested How would we make it or join this research project? please feel free to collaborate with us!

In this paper, we propose an explainable visual dataset, XIMAGENET-12, to evaluate the robustness of visual models. XIMAGENET-12 consists of over 200K images with 15,410 manual semantic annotations. Specifically, we deliberately selected 12 categories from ImageNet, representing objects commonly encountered in practical life. To simulate real-world situations, we incorporated six diverse scenarios, such as overexposure, blurring, and color changes, etc. We further develop a quantitative criterion for robustness assessment, allowing for a nuanced understanding of how visual models perform under varying conditions, notably in relation to the background.

Progress:

• Blur Background-> Done! You can find the image Generated in the corresponding folder!
• Segmented Background -> Done! you can download the image and its corresponding transparent mask image!
• Color in Background->Done!~~ you can now download the image with different background color modified, and play with different color-ed images!
• Random Background with Real Environment -> Done! you can also find we generated the image with the photographer's real image as a background and removed the original background of the target object, but similar to the style!
• Bias of tools during annotation->Done! for this one, you won't get a new image, because this is about math and statistics data analysis when different tools and annotators are applied!
• AI generated Background-> current on progress ( 12 /12) Done!, So basically you can find one sample folder image we uploaded, please take a look at how real it is, and guess what LLM model we are using to generate the high-resolution background to make it so real :)

What tool we used to generate those images?

We employed a combination of tools and methodologies to generate the images in this dataset, ensuring both efficiency and quality in the annotation and synthesis processes.

• IoG Net: Initially, we utilized the IoG Net, which played a foundational role in our image generation pipeline.
• Polygon Faster Labeling Tool: To facilitate the annotation process, we developed a custom Polygon Faster Labeling Tool, streamlining the labeling of objects within the images.AnyLabeling Open-source Project: We also experimented with the AnyLabeling open-source project, exploring its potential for our annotation needs.
• V7 Lab Tool: Eventually, we found that the V7 Lab Tool provided the most efficient labeling speed and delivered high-quality annotations. As a result, we standardized the annotation process using this tool.
• Data Augmentation: For the synthesis of synthetic images, we relied on a combination of deep learning frameworks, including scikit-learn and OpenCV. These tools allowed us to augment and manipulate images effectively to create a diverse range of backgrounds and variations.
• GenAI: Our dataset includes images generated using the Stable Diffusion XL model, along with versions 1.5 and 2.0 of the Stable Diffusion model. These generative models played a pivotal role in crafting realistic and varied backgrounds.

For a detailed breakdown of our prompt engineering and hyperparameters, we invite you to consult our upcoming paper. This publication will provide comprehensive insights into our methodologies, enabling a deeper understanding of the image generation process.

How to use our dataset?

this dataset has been/could be downloaded via Kaggl...

Community science image libraries offer a massive, but largely untapped, source of observational data for phenological research. The iNaturalist platform offers a particularly rich archive, containing more than 49 million verifiable, georeferenced, open access images, encompassing seven continents and over 278,000 species. A critical limitation preventing scientists from taking full advantage of this rich data source is labor. Each image must be manually inspected and categorized by phenophase, which is both time-intensive and costly. Consequently, researchers may only be able to use a subset of the total number of images available in the database. While iNaturalist has the potential to yield enough data for high-resolution and spatially extensive studies, it requires more efficient tools for phenological data extraction. A promising solution is automation of the image annotation process using deep learning. Recent innovations in deep learning have made these open-source tools accessible to a general research audience. However, it is unknown whether deep learning tools can accurately and efficiently annotate phenophases in community science images. Here, we train a convolutional neural network (CNN) to annotate images of Alliaria petiolata into distinct phenophases from iNaturalist and compare the performance of the model with non-expert human annotators. We demonstrate that researchers can successfully employ deep learning techniques to extract phenological information from community science images. A CNN classified two-stage phenology (flowering and non-flowering) with 95.9% accuracy and classified four-stage phenology (vegetative, budding, flowering, and fruiting) with 86.4% accuracy. The overall accuracy of the CNN did not differ from humans (p = 0.383), although performance varied across phenophases. We found that a primary challenge of using deep learning for image annotation was not related to the model itself, but instead in the quality of the community science images. Up to 4% of A. petiolata images in iNaturalist were taken from an improper distance, were physically manipulated, or were digitally altered, which limited both human and machine annotators in accurately classifying phenology. Thus, we provide a list of photography guidelines that could be included in community science platforms to inform community scientists in the best practices for creating images that facilitate phenological analysis.

Methods Creating a training and validation image set

We downloaded 40,761 research-grade observations of A. petiolata from iNaturalist, ranging from 1995 to 2020. Observations on the iNaturalist platform are considered “research-grade if the observation is verifiable (includes image), includes the date and location observed, is growing wild (i.e. not cultivated), and at least two-thirds of community users agree on the species identification. From this dataset, we used a subset of images for model training. The total number of observations in the iNaturalist dataset are heavily skewed towards more recent years. Less than 5% of the images we downloaded (n=1,790) were uploaded between 1995-2016, while over 50% of the images were uploaded in 2020. To mitigate temporal bias, we used all available images between the years 1995 and 2016 and we randomly selected images uploaded between 2017-2020. We restricted the number of randomly-selected images in 2020 by capping the number of 2020 images to approximately the number of 2019 observations in the training set. The annotated observation records are available in the supplement (supplementary data sheet 1). The majority of the unprocessed records (those which hold a CC-BY-NC license) are also available on GBIF.org (2021).

One of us (R. Reeb) annotated the phenology of training and validation set images using two different classification schemes: two-stage (non-flowering, flowering) and four-stage (vegetative, budding, flowering, fruiting). For the two-stage scheme, we classified 12,277 images and designated images as ‘flowering’ if there was one or more open flowers on the plant. All other images were classified as non-flowering. For the four-stage scheme, we classified 12,758 images. We classified images as ‘vegetative’ if no reproductive parts were present, ‘budding’ if one or more unopened flower buds were present, ‘flowering’ if at least one opened flower was present, and ‘fruiting’ if at least one fully-formed fruit was present (with no remaining flower petals attached at the base). Phenology categories were discrete; if there was more than one type of reproductive organ on the plant, the image was labeled based on the latest phenophase (e.g. if both flowers and fruits were present, the image was classified as fruiting).

For both classification schemes, we only included images in the model training and validation dataset if the image contained one or more plants with clearly visible reproductive parts were clear and we could exclude the possibility of a later phenophase. We removed 1.6% of images from the two-stage dataset that did not meet this requirement, leaving us with a total of 12,077 images, and 4.0% of the images from the four-stage leaving us with a total of 12,237 images. We then split the two-stage and four-stage datasets into a model training dataset (80% of each dataset) and a validation dataset (20% of each dataset).

Training a two-stage and four-stage CNN

We adapted techniques from studies applying machine learning to herbarium specimens for use with community science images (Lorieul et al. 2019; Pearson et al. 2020). We used transfer learning to speed up training of the model and reduce the size requirements for our labeled dataset. This approach uses a model that has been pre-trained using a large dataset and so is already competent at basic tasks such as detecting lines and shapes in images. We trained a neural network (ResNet-18) using the Pytorch machine learning library (Psake et al. 2019) within Python. We chose the ResNet-18 neural network because it had fewer convolutional layers and thus was less computationally intensive than pre-trained neural networks with more layers. In early testing we reached desired accuracy with the two-stage model using ResNet-18. ResNet-18 was pre-trained using the ImageNet dataset, which has 1,281,167 images for training (Deng et al. 2009). We utilized default parameters for batch size (4), learning rate (0.001), optimizer (stochastic gradient descent), and loss function (cross entropy loss). Because this led to satisfactory performance, we did not further investigate hyperparameters.

Because the ImageNet dataset has 1,000 classes while our data was labeled with either 2 or 4 classes, we replaced the final fully-connected layer of the ResNet-18 architecture with fully-connected layers containing an output size of 2 for the 2-class problem and 4 for the 4-class problem. We resized and cropped the images to fit ResNet’s input size of 224x224 pixels and normalized the distribution of the RGB values in each image to a mean of zero and a standard deviation of one, to simplify model calculations. During training, the CNN makes predictions on the labeled data from the training set and calculates a loss parameter that quantifies the model’s inaccuracy. The slope of the loss in relation to model parameters is found and then the model parameters are updated to minimize the loss value. After this training step, model performance is estimated by making predictions on the validation dataset. The model is not updated during this process, so that the validation data remains ‘unseen’ by the model (Rawat and Wang 2017; Tetko et al. 1995). This cycle is repeated until the desired level of accuracy is reached. We trained our model for 25 of these cycles, or epochs. We stopped training at 25 epochs to prevent overfitting, where the model becomes trained too specifically for the training images and begins to lose accuracy on images in the validation dataset (Tetko et al. 1995).

We evaluated model accuracy and created confusion matrices using the model’s predictions on the labeled validation data. This allowed us to evaluate the model’s accuracy and which specific categories are the most difficult for the model to distinguish. For using the model to make phenology predictions on the full, 40,761 image dataset, we created a custom dataloader function in Pytorch using the Custom Dataset function, which would allow for loading images listed in a csv and passing them through the model associated with unique image IDs.

Hardware information

Model training was conducted using a personal laptop (Ryzen 5 3500U cpu and 8 GB of memory) and a desktop computer (Ryzen 5 3600 cpu, NVIDIA RTX 3070 GPU and 16 GB of memory).

Comparing CNN accuracy to human annotation accuracy

We compared the accuracy of the trained CNN to the accuracy of seven inexperienced human scorers annotating a random subsample of 250 images from the full, 40,761 image dataset. An expert annotator (R. Reeb, who has over a year’s experience in annotating A. petiolata phenology) first classified the subsample images using the four-stage phenology classification scheme (vegetative, budding, flowering, fruiting). Nine images could not be classified for phenology and were removed. Next, seven non-expert annotators classified the 241 subsample images using an identical protocol. This group represented a variety of different levels of familiarity with A. petiolata phenology, ranging from no research experience to extensive research experience (two or more years working with this species). However, no one in the group had substantial experience classifying community science images and all were naïve to the four-stage phenology scoring protocol. The trained CNN was also used to classify the subsample images. We compared human annotation accuracy in each phenophase to the accuracy of the CNN using students

Open-Source Urdu OCR Dataset & Baseline

Overview

This repository/dataset contains 8,000+ images of Urdu text along with corresponding ground truth text files. The data includes both printed and handwritten Urdu samples, aiming to support the development of a high-quality OCR system for Urdu.

Urdu is spoken by over 100 million people worldwide but lacks robust open-source OCR tools. This dataset and project provide the foundation to build and improve Urdu OCR models using modern deep learning techniques. Dataset Details

Number of images: 8,000+

Image types: Scanned printed text, handwritten text, and natural scene Urdu text (where available)

Annotations: Text transcripts aligned at the line or paragraph level

Format: Images (PNG/JPEG), ground truth text files (TXT) with matching filenames

License: MIT

Getting Started Download

Access the dataset here: [https://www.kaggle.com/datasets/humairmunir/ocr-dataset-8-k-images-and-gt-text] Baseline Models

We provide starter notebooks demonstrating:

Basic data loading and preprocessing

Training a simple CRNN or Transformer-based OCR model

Evaluating performance with Character Error Rate (CER) and Word Error Rate (WER)

Notebooks: [Insert links to starter notebooks] How to Contribute

We welcome contributions in the following areas:

Dataset expansion: Add more scanned or photographed Urdu text images, handwritten notes, or scene text.

Annotation: Help with aligning images and transcriptions, bounding box labeling.

Model development: Share improved OCR models, preprocessing pipelines, and post-processing techniques.

Evaluation: Propose new metrics or build benchmarking tools.

Contact & Collaboration

If you have data or want to contribute, please reach out: 📧 humairmunirawan@gmail.com

You can also share datasets or code via Google Drive, Dropbox, or GitHub pull requests. Why Urdu OCR?

Urdu has millions of speakers but limited digital resources. This project aims to close that gap by creating accessible, accurate OCR tools, supporting digitization, research, and technology innovation.

License

MIT

Feel free to ask if you want me to generate:

A starter notebook template

Annotation guidelines

Let me know!

Community science image libraries offer a massive, but largely untapped, source of observational data for phenological research. The iNaturalist platform offers a particularly rich archive, containing more than 49 million verifiable, georeferenced, open access images, encompassing seven continents and over 278,000 species. A critical limitation preventing scientists from taking full advantage of this rich data source is labor. Each image must be manually inspected and categorized by phenophase, which is both time-intensive and costly. Consequently, researchers may only be able to use a subset of the total number of images available in the database. While iNaturalist has the potential to yield enough data for high-resolution and spatially extensive studies, it requires more efficient tools for phenological data extraction. A promising solution is automation of the image annotation process using deep learning. Recent innovations in deep learning have made these open-source tools accessible to a general research audience. However, it is unknown whether deep learning tools can accurately and efficiently annotate phenophases in community science images. Here, we train a convolutional neural network (CNN) to annotate images of Alliaria petiolata into distinct phenophases from iNaturalist and compare the performance of the model with non-expert human annotators. We demonstrate that researchers can successfully employ deep learning techniques to extract phenological information from community science images. A CNN classified two-stage phenology (flowering and non-flowering) with 95.9% accuracy and classified four-stage phenology (vegetative, budding, flowering, and fruiting) with 86.4% accuracy. The overall accuracy of the CNN did not differ from humans (p = 0.383), although performance varied across phenophases. We found that a primary challenge of using deep learning for image annotation was not related to the model itself, but instead in the quality of the community science images. Up to 4% of A. petiolata images in iNaturalist were taken from an improper distance, were physically manipulated, or were digitally altered, which limited both human and machine annotators in accurately classifying phenology. Thus, we provide a list of photography guidelines that could be included in community science platforms to inform community scientists in the best practices for creating images that facilitate phenological analysis.