This paper tackles a novel problem: how to transfer knowledge from the emerging Segment Anything Model (SAM) to learn a compact panoramic semantic segmentation model, i.e., student, without requiring any labeled data.

Dataset Title: Bugzz lightyears: To Semantic Segmentation and Bug-yond!

Description:

This dataset comprises a collection of real and robotic toy bugs designed for a small-scale semantic segmentation project. Each bug has been captured six times from various angles, ensuring comprehensive coverage of their features and details. The dataset serves as a valuable resource for exploring semantic segmentation techniques and evaluating machine learning models.

Dataset Details:

• Images: Each bug is represented by six images taken from different perspectives, facilitating robust segmentation and analysis.
• Segmentation: The dataset has been meticulously segmented using Label Studio in conjunction with the SAM (Segment Anything Model), enabling precise delineation of each bug from the background.
• Diversity: The collection includes a variety of bugs, both real and robotic, providing a unique blend for training and testing segmentation models.

Usage: This toy dataset is ideal for researchers and developers interested in:

• Experimenting with semantic segmentation algorithms.
• Developing and refining computer vision models for object detection and segmentation.
• Educational purposes in machine learning and computer vision courses.

License: This dataset is made available under [specify license type, e.g., CC BY 4.0], allowing for both academic and commercial use, with proper attribution to the creator.

The dataset is an extension of the Semantic Drone Dataset of Institute of Computer Graphics and Vision at the Graz University of Technology.

Binary and 5-class Extension

The extension proposes two different preprocessed datasets in order to perform binary segmentation and multi-class segmentation with 5 macro-groups instead of the original 24 labels and a resolution of 960x736px instead of 6000x4000px.

Colormaps and Re-labeling

All the information relative to the colors assigned to each class are contained in the colormaps.xlsx file and in addition to it there are also the conversion dictionaries used to convert the labels in classes_dict.txt.

Original Dataset

The original dataset with 24 different classes and 24Mpx of resolution is contained in the folder semantic drone dataset

Questions and Comments

Leave an up-vote if you are going to use this dataset or leave a comment/suggestion on how I could improve the documentation, if you have questions feel free to ask

These figures are the graphical results of my Master 2 internship on automatic segmentation using SAM2(Segment Anything Model 2)an artificial intelligence. The red line represents the best cell line from which anatomical measurements were made.

COCO semantic segmentation maps

This dataset contains semantic segmentation maps (monochrome images where each pixel corresponds to one of the 133 COCO categories used for panoptic segmentation). It was generated from the 2017 validation annotations using the following process:

git clone https://github.com/cocodataset/panopticapi and install it. python converters/panoptic2semantic_segmentation.py --input_json_file /data/datasets/coco/2017/annotations/panoptic_val2017.json… See the full description on the dataset page: https://huggingface.co/datasets/enterprise-explorers/coco-semantic-segmentation.

Visual comparison of 100 human annotations (labels) compared with Segment Anything Model 2 (SAM2) segmentation.

The Segment Anything Model (SAM) has been proven to be a powerful foundation model for image segmentation tasks, which is an important task in computer vision. However, the transfer of its rich semantic information to multiple different downstream tasks remains unexplored. In this paper, we propose the Task-Aware Low-Rank Adaptation (TA-LoRA) method, which enables SAM to work as a foundation model for multi-task learning.

This dataset images are collected from tropical Malaysian forests and encompasses a diverse range of arthropod species captured under various lighting and environmental conditions. There are 4,949 original images and 4,949 corresponding segmentation masks in the dataset. The dataset images contain 2 classes, namely background and foreground; so this task can be considered as binary semantic segmentation task.

The structure of the data is as follows:

• ROOT -images: - img_file; - img_file; - img_file; - ........ - img_file.

  -labels: - img_file; - img_file; - img_file; - ........ - img_file.

The images in the dataset have various resolutions; thus, they must be resized before training process. Good luck!

Sample Semantic Segmentation

## Overview

Sample Semantic Segmentation is a dataset for semantic segmentation tasks - it contains Veggies annotations for 6,331 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [MIT license](https://creativecommons.org/licenses/MIT).

MIT Indoor Semantic Segmentation

## Overview

MIT Indoor Semantic Segmentation is a dataset for semantic segmentation tasks - it contains Indoor Objects annotations for 2,582 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

This dataset contains 10 examples of the segments/sidewalk-semantic dataset (i.e. 10 images with corresponding ground-truth segmentation maps).

A collection of 4,949 original insect images and 4,949 segmented versions captured in Malaysian forests. The dataset supports semantic segmentation tasks for agriculture, environmental science, entomology, and biodiversity studies.

Semantic Segmentation AI Market Outlook

According to our latest research, the global Semantic Segmentation AI market size reached USD 2.14 billion in 2024, with a robust compound annual growth rate (CAGR) of 24.6% expected from 2025 to 2033. By the end of the forecast period, the market is projected to attain a value of USD 16.18 billion by 2033. This remarkable growth is primarily driven by the increasing adoption of artificial intelligence in image and video analysis across diverse industries, including healthcare, automotive, and manufacturing. The market’s expansion is further supported by advancements in deep learning algorithms and the proliferation of high-resolution imaging devices, which together enhance the accuracy and efficiency of semantic segmentation solutions.

The surge in demand for automated systems in sectors such as autonomous vehicles, medical diagnostics, and industrial automation is a significant growth factor for the Semantic Segmentation AI market. With the rapid evolution of computer vision technologies, businesses are leveraging semantic segmentation to extract meaningful insights from visual data, enabling improved decision-making and operational efficiency. The integration of AI-driven segmentation in autonomous vehicles, for example, is critical for real-time object detection and scene understanding, which directly contributes to enhanced safety and navigation capabilities. Similarly, in healthcare, semantic segmentation is revolutionizing medical imaging by enabling precise identification of anatomical structures and pathological regions, thereby improving diagnostic accuracy and patient outcomes.

Another major driver fueling the growth of the Semantic Segmentation AI market is the increasing deployment of AI-powered surveillance and security systems. The growing need for advanced monitoring solutions in urban infrastructure, public safety, and critical facilities has led to a surge in demand for real-time semantic understanding of video feeds. This trend is further amplified by the proliferation of smart cities and the adoption of Internet of Things (IoT) devices, which generate vast amounts of visual data requiring efficient processing and analysis. As organizations strive to enhance situational awareness and threat detection, semantic segmentation AI is emerging as a vital tool for delivering actionable intelligence from complex visual environments.

Furthermore, the market is witnessing significant investments in research and development, aimed at improving the scalability, accuracy, and computational efficiency of semantic segmentation algorithms. The advent of edge computing and the increasing availability of high-performance hardware are enabling the deployment of AI models closer to the data source, reducing latency and bandwidth requirements. This shift towards edge-based processing is particularly beneficial in applications such as robotics and agriculture, where real-time decision-making is crucial. As a result, the Semantic Segmentation AI market is poised for sustained growth, driven by technological innovations and the expanding scope of AI applications across traditional and emerging sectors.

From a regional perspective, North America currently leads the Semantic Segmentation AI market, accounting for the largest share in 2024, followed closely by Europe and Asia Pacific. The presence of major technology players, robust research infrastructure, and early adoption of AI technologies are key factors underpinning North America’s dominance. Meanwhile, Asia Pacific is expected to exhibit the highest CAGR during the forecast period, propelled by rapid industrialization, increasing investments in AI research, and a burgeoning ecosystem of startups. Europe also demonstrates strong growth potential, driven by advancements in automotive AI and smart manufacturing initiatives. Latin America and the Middle East & Africa, though smaller in market size, are gradually embracing AI-powered segmentation solutions, particularly in surveillance and agricultural applications.

Component Analy

The performance of different semantic segmentation models on the self-constructed training dataset.

The authors of the ADE20K dataset address the significant challenge of scene parsing, encompassing the recognition and segmentation of objects and stuff within images, a vital task in the domain of computer vision. Despite the efforts made by the research community to gather data, there remains a scarcity of image datasets that comprehensively cover a broad spectrum of scenes and object categories, along with detailed and dense annotations suitable for scene parsing. To fill this void, the authors introduce the ADE20K dataset. This dataset features diverse annotations that span scenes, objects, parts of objects, and, intriguingly, even parts of parts. In order to facilitate benchmarking for scene parsing, the ADE20K dataset includes 150 object and stuff classes, and various segmentation baseline models undergo evaluation using this benchmark. You can access the hierarchy of classes on the official website of the dataset.

EVA

https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F19272950%2F3136c2a234771726dd17c29a758ba365%2Fb0.png?generation=1709156580593893&alt=media" alt="">

Fig. 1: Diagram of the proposed blueberry fruit phenotyping workflow, involving four stages: data collection, dataset generation, model training, and phenotyping traits extraction. Our mobile platform equipped with a multi-view imaging system (top, left and right) was used to scan the blueberry plants through navigating over crop rows. On the basis of fruit/cluster detection dataset, we leverage a maturity classifier and a segmentation foundation model, SAM, to generate a semantic instance dataset for immature, semi-mature, and mature fruits segmentation. We proposed a lightweight improved YOLOv8 model for fruit cluster detection and blueberry segmentation for plant-scale and cluster-scale phenotyping traits extraction, including yield, maturity, cluster number and compactness.

Dataset generation: https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F19272950%2F7a06785e03056ac75a41f0ba881c7ca2%2Fb1.png?generation=1709156618386382&alt=media" alt=""> Fig 2: Illumination of the proposed automated pixel-wise labels generation for immature, semi-mature, and mature blueberry fruits (genotype: keecrisp). From left to right: (a) bounding box labels of blueberries from our previous manual detection dataset [27]; (b) three-classes boxes labels (immature-yellow, semi-mature-red, mature-blue) re-classified with a maturity classifier; (c) pixel-wise mask labels of blueberry fruits with Segment Anything Model.

References

If you find this work or code useful, please cite:

@article{li2025-robotic blueberry phenotyping,
 title={In-field blueberry fruit phenotyping with a MARS-PhenoBot and customized BerryNet},
 author={Li, Zhengkun and Xu, Rui and Li, Changying and Munoz, Patricio and Takeda, Fumiomi and Leme, Bruno},
 journal={Computers and Electronics in Agriculture},
 volume={232},
 pages={110057},
 year={2025},
 publisher={Elsevier}
}

The performance of different semantic segmentation models on the self-constructed validation dataset.

Performance evaluation of semantic segmentation models is an essential task because it helps to identify the best-performing model. Traditional methods, however, are generally concerned with the improvement of a single quality or quantity. Moreover, what causes low performance usually goes unnoticed. To address these issues, a new cross meta-frontier data envelopment analysis (DEA) approach is proposed in this article. For evaluating model performance comprehensively, not only accuracy metrics, but also hardware burden and model structure factors, are taken as DEA outputs and inputs, separately. In addition, the potential inefficiency is attributed to architectures and backbones via efficiency decomposition, so that it can find the sources of inefficiency and provides a direction for performance improvement. Finally, based on the proposed approach, the performance of 16 classical semantic segmentation models on the PASCAL VOC dataset are re-evaluated and explained. The results verify that the proposed approach can be considered as a comprehensive and interpretable performance evaluation technique, which expands the traditional accuracy-based measurement.

The dataset used for the experiments with the proposed approach to augment image data for semantic segmentation networks by applying image-to-image translation with both, a domain-specific mathematical model and an approach entirely based on generative models.