The authors of the PASCAL Context dataset conduct a comprehensive investigation into the significance of context within existing state-of-the-art detection and segmentation methodologies. Their approach involves the meticulous labeling of every pixel encompassed within the PASCAL VOC 2010 detection challenge, associating each pixel with a semantic category. This dataset is envisioned to present a considerable challenge to the research community, as it incorporates an impressive 520 additional classes that cater to both semantic segmentation and object detection.

This dataset is a set of additional annotations for PASCAL VOC 2010. It goes beyond the original PASCAL semantic segmentation task by providing annotations for the whole scene. The statistics section has a full list of 400+ labels. Every pixel has a unique class label. Instance information (i.e, different masks to separate different instances of the same class in the same image) are currently provided for the 20 PASCAL objects. Statistics Since the dataset is an annotation of PASCAL VOC 2010, it has the same statistics as those of the original dataset. Training and validation contains 10,103 images while testing contains 9,637 images. Usage Considerations The classes are not drawn from a fixed pool. Instead labelers were free to either select or type in what they believe to be the appropriate class and to determine what the appropriate object granularity is. We decided to merge/split some of the categories so the current number of categories is different from what we mentioned in the C

COCO dataset, ADE20K dataset, PASCAL Context dataset, LVIS dataset

Dataset

This dataset was created by Adel Samigullin

Contents

Dataset

This dataset was created by Fatemeh Boloori

Released under Apache 2.0

Contents

The PASCAL VOC project:

• Provides standardised image data sets for object class recognition
• Provides a common set of tools for accessing the data sets and annotations
• Enables evaluation and comparison of different methods
• Ran challenges evaluating performance on object class recognition (from 2005-2012, now finished)

Context

The goal of this challenge is to recognize objects from a number of visual object classes in realistic scenes (i.e. not pre-segmented objects). It is fundamentally a supervised learning learning problem in that a training set of labelled images is provided. The twenty object classes that have been selected are:

• Person: person
• Animal: bird, cat, cow, dog, horse, sheep
• Vehicle: aeroplane, bicycle, boat, bus, car, motorbike, train
• Indoor: bottle, chair, dining table, potted plant, sofa, tv/monitor
• There will be two main competitions, and two smaller scale "taster" competitions.

Content

The training data provided consists of a set of images; each image has an annotation file giving a bounding box and object class label for each object in one of the twenty classes present in the image. Note that multiple objects from multiple classes may be present in the same image.

Acknowledgements

@misc{pascal-voc-2007, author = "Everingham, M. and Van~Gool, L. and Williams, C. K. I. and Winn, J. and Zisserman, A.", title = "The {PASCAL} {V}isual {O}bject {C}lasses {C}hallenge 2007 {(VOC2007)} {R}esults", howpublished = "http://www.pascal-network.org/challenges/VOC/voc2007/workshop/index.html"}

In the context of intelligent driving, pedestrian detection faces challenges related to low accuracy in target recognition and positioning. To address this issue, a pedestrian detection algorithm is proposed that integrates a large kernel attention mechanism with the YOLOV5 lightweight model. The algorithm aims to enhance long-term attention and dependence during image processing by fusing the large kernel attention module with the C3 module. Furthermore, it addresses the lack of long-distance relationship information in channel and spatial feature extraction and representation by introducing the Coordinate Attention mechanism. This mechanism effectively extracts local information and focused location details, thereby improving detection accuracy. To improve the positioning accuracy of obscured targets, the alpha CIOU bounding box regression loss function is employed. It helps mitigate the impact of occlusions and enhances the algorithm’s ability to precisely localize pedestrians. To evaluate the effectiveness of trained model, experiments are conducted on the BDD100K pedestrian dataset as well as the Pascal VOC dataset. Experimental results demonstrate that the improved attention fusion YOLOV5 lightweight model achieves an average accuracy of 60.3%. Specifically, the detection accuracy improves by 1.1% compared to the original YOLOV5 algorithm, and the accuracy performance index reaches 73.0%. These findings strongly indicate the proposed algorithm in significantly enhancing the accuracy of pedestrian detection in road scenes.

Context

Original Dataset is used in the paper, (S. Liu, J. Feng, C. Domokos, H. Xu, J. Huang, Z. Hu, & S. Yan. 2014) CFPD | Fashion Parsing with Weak Color-Category Labels, with for Object Detection and Segmentation tasks (https://sites.google.com/site/fashionparsing)

This dataset is custom for Object Detection task, with remove skin, face, background infomation, and format follow PASCAL VOC format. The classes of the this dataset: -sunglass, -hat, -jacket, -shirt, -pants, -shorts, -skirt, -dress, -bag, -shoe

Note: If you want .txt file with YOLO format, you can use Annotations_txt directory.

This dataset was collected within the context of the PAsCAL research project between January 2022 and February 2022 at the ACI Vallelunga test circuit and premises in Rome, Italy. Subject of the pilot was a driving training for advanced ADAS systems and test driving of a Level-2+ autonomous vehicle on a test track, performing several different manoeuvres to test the capability of the ADAS systems.

Some of the participants were subjected to a driving training for autonomous vehicles before they did the test drive, wherein they had to perform several difficult driving manoeuvres (such as on slippery ground or emergency braking). The purpose of this pilot was to observe whether a driving training improves the driver's capability to use ADAS systems and therefore operate the vehicle in a safer way. Depending on the pilot, it is recommended to adapt existing driving training for beginners, professionals and experienced drivers.

In order to analyse the answers given to the questions, it is recommended to consult also the "PAsCAL WP6 Pilots Surveys" dataset, which contains all questions and possible answers.

These two datasets were collected within the context of the PAsCAL research project between September 2021 and March 2022 on the campus of the UAM University in Madrid, Spain. Subject of the pilot was a Level 4 autonomous bus shuttle, which is to date one of the only shuttles in Europe to run in open traffic. Due to this and the fact that only a steward is on-board of the vehicle in case of incidences or passenger support, two surveys were designed:

Survey for Shuttle Users: Passengers experienced the ride on the autonomous shuttle within the context of the multi-modal trip, connecting them to an interurban train station and an interurban (long-distance) bus station on the other side. Purpose of the survey was to capture the participant's overall acceptance and attitude towards the vehicle after using it and comparing it directly to available traditional modes of transport.

Survey for Shuttle Co-Road Users: Since the shuttle is operating in open traffic, co-road users were also stopped randomly and asked to complete the survey to map the acceptance of the autonomous shared and public vehicle they were sharing the road with. This included not just car drivers, but also pedestrians and cyclists on-site.

In order to analyse the answers given to the questions, it is recommended to consult also the "PAsCAL WP6 Pilots Surveys" dataset, which contains all questions and possible answers.

Contrast of detection results of different algorithms in PASCAL VOC2007 dataset.

Context

This dataset was created to show, how one can create an object detector from scratch without writing any code. More about that here https://youtu.be/wuVh1X-HbJ8

Content

The dataset contains sheep images, which were collected from the Internet and annotated by VoTT visual object tagging tool. More about that here https://youtu.be/uDWgWJ5Gpwc

Acknowledgement

This dataset was created by Intelec AI team.

This dataset was collected within the context of the PAsCAL research project between June 2021 and September 2021 at the E-Bus Competence Centre premises in Livange, Luxembourg. Subject of the pilot was a Wizard of Oz experiment, which consisted of a modified high-capacity bus vehicle to simulate a Level-5 automated vehicle to the passengers, although the bus was operated and driven by a human driver, which was not visible to the participants. The participants experienced several malfunctions of the bus and were offered a Human-Machine-Interface (HMI), which connected them to a traffic control centre for troubleshooting. The purpose of the pilot was to evaluate whether available HMIs are able to bridge the gap to human driver support to passengers. Some of the participants were blind or partially sighted to also observe the adequacy of the solution on vulnerable travellers. In order to analyse the answers given to the questions, it is recommended to consult also the "PAsCAL WP6 Pilots Surveys" dataset, which contains all questions and possible answers.

The comparison experimental results (pixel accuracy, mean accuracy, and mean IoU) of the SIEANs with different methods on Stanford Background dataset.

These two datasets were collected within the context of the PAsCAL research project between November 2021 and June 2022. Each of the surveys were dedicated to two different pilot scenarios, settings and vehicles:

Shared Connected Vehicle Fleet: An existing rental service for a vehicle fleet for employees and students of the University of Luxembourg was enhanced by adding an advances Level-2+ vehicle to the fleet. The users were already familiar with the functionality of the service (booking process, etc.) and were asked to take a realistic trip including urban areas but also a strip of highway and were invited to test the autonomous features of the vehicle (removing hands from steering wheel, automatic parking and many more). The pilot took place in the Belval area of Luxembourg;

Bus shuttle: An autonomous bus shuttle with Level-4 autonomy was piloted, which connects a train station to a business park. Participants were workers or visitors of the business park and the objective of this pilot was to observe the adequacy of the shuttle in the commuting context.

In order to analyse the answers given to the questions, it is recommended to consult also the "PAsCAL WP6 Pilots Surveys" dataset, which contains all questions and possible answers.

We evaluated our AHOD model using two well-known datasets in the field of object detection:COCO (Common Objects in Context)One of the most widely used benchmarks for object detection.Contains over 200,000 images and more than 80 object categories.Includes objects in varied and sometimes cluttered contexts, allowing the robustness of detectors to be evaluated.Pascal VOCAnother reference dataset, often used for classification, detection and segmentation tasks.Includes 20 object categories, with precise bounding box annotations.Less complex than COCO, but useful for comparing performance on more conventional objects.Tools, techniques and innovations usedThe AHOD architecture is based on three main modules:Feature Pyramid Enhancement (FPE)Multi-scale feature processing tool.Improves the representation of objects of various sizes in the same image.Inspired by architectures such as FPN (Feature Pyramid Networks), but optimised for better performance.Dynamic Context Module (DCM)Intelligent contextual module.Capable of dynamically adjusting the extracted features according to the context (e.g. by adapting the features according to urban or rural areas in a road image).Enhances the model's ability to understand the overall context of the scene.Fast and Accurate Detection Head (FADH)Optimised detection head.Seeks a compromise between the speed of YOLO and the accuracy of Faster R-CNN.Probably uses lightweight convolution layers or optimisations such as MobileNet/Depthwise Convolutions.Probable technologies usedAlthough the summary does not specify this, we can reasonably assume that the following tools are used:Deep learning frameworks: PyTorch or TensorFlow, which are standard in object detection research.GPUs for training and inference, particularly for measuring inference times (essential in real-time applications).Standard evaluation techniques:mAP (mean Average Precision): measure of average precision.FPS (Frames Per Second) or inference time for real-time performance.

Small object detection is an essential but challenging task in computer vision. Transformer-based algorithms have demonstrated remarkable performance in the domain of computer vision tasks. Nevertheless, they suffer from inadequate feature extraction for small objects. Additionally, they face difficulties in deployment on resource-constrained platforms due to their heavy computational burden. To tackle these problems, an efficient local-global fusion Transformer (ELFT) is proposed for small object detection, which is based on attention and grouping strategy. Specifically, we first design an efficient local-global fusion attention (ELGFA) mechanism to extract sufficient location features and integrate detailed information from feature maps, thereby promoting the accuracy. Besides, we present a grouped feature update module (GFUM) to reduce computational complexity by alternately updating high-level and low-level features within each group. Furthermore, the broadcast context module (CB) is introduced to obtain richer context information. It further enhances the ability to detect small objects. Extensive experiments conducted on three benchmarks, i.e. Remote Sensing Object Detection (RSOD), NWPU VHR-10 and PASCAL VOC2007, achieving 95.8%, 94.3% and 85.2% in mean average precision (mAP), respectively. Compared to DINO, the number of parameters is reduced by 10.4%, and the floating point operations (FLOPs) are reduced by 22.7%. The experimental results demonstrate the efficacy of ELFT in small object detection tasks, while maintaining an attractive level of computational complexity.

The Stanford Background Dataset is a new dataset introduced in Gould et al. (ICCV 2009) for evaluating methods for geometric and semantic scene understanding. The dataset contains 715 images chosen from existing public datasets: LabelMe, MSRC, PASCAL VOC and Geometric Context. Our selection criteria were for the images to be of outdoor scenes, have approximately 320-by-240 pixels, contain at least one foreground object, and have the horizon position within the image (it need not be visible).

Comparison of detection results of different algorithms in NWPU VHR-10 dataset.

The results of ablation studies on the RSOD dataset.