Introduction #
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.
Note, that there are 42 unlabeled classes in the dataset.
The authors delve into the role of context in perceptual inference, emphasizing the natural proficiency of humans in comprehending the visual world. Context serves as a pivotal statistical feature of the world, enhancing the efficiency and precision of perceptual inference tasks. Previous cognition-based studies have underscored the role of context in diverse perceptual tasks, including object detection, semantic segmentation, and scene classification. The authors emphasize the influence of contextual information, such as object arrangements in specific scenes, relative object size, and location, on human object detection. Additionally, the authors cite studies demonstrating that humans exhibit superior performance in perceptual tasks when contextual information is available, including object segmentation and image patch classification. The authors aim to delve further into the effect of context within detection and segmentation approaches, setting the stage for their investigation by meticulously labeling every pixel within the training and validation sets of the PASCAL VOC 2010 detection challenge with a semantic class. PASCAL is selected as the testbed due to its prominent role as a benchmark for detection and segmentation in the research community.
The authors’ analysis substantiates the heightened challenge of their dataset compared to alternatives, attributing this challenge to factors such as higher class entropy and a greater diversity of object categories. Their dataset features pixel-wise labels for the 10K trainval images from the PASCAL VOC 2010 detection challenge, encompassing 540 categories divided into objects, stuff, and hybrid classes. The painstaking annotation process spanned three months and involved six in-house annotators, resulting in highly accurate segmentations. These annotations surpass the quality of those generated through online systems like MTurk. Annotators were presented with an initial set of 80 labels and encouraged to introduce additional classes for cases not covered by the initial set. In instances of ambiguity, regions were labeled as “unknown.” Each annotation underwent a double-check, with necessary revisions to ensure coherence. Following a power law distribution, the most frequent 59 classes were selected for analysis, while the remaining classes were assigned the background label. Notably, the vast majority of pixels, approximately 87.2%, were labeled as foreground, contrasting with only 29.3% of pixels covered by the 20 object classes within PASCAL VOC. A distribution of pixels and images among the 59 most frequent categories is visually presented:
Summary #
PASCAL Context Dataset is a dataset for a semantic segmentation task. It is applicable or relevant across various domains.
The dataset consists of 10103 images with 65937 labeled objects belonging to 459 different classes including person, unknown, tree, and other: sky, building, ground, wall, grass, floor, table, pole, chair, car, dog, road, fence, cat, light, water, cloth, sofa, bottle, window, mountain, picture, bird, pottedplant, aeroplane, and 431 more.
Images in the PASCAL Context dataset have pixel-level semantic segmentation annotations. All images are labeled (i.e. with annotations). There are 2 splits in the dataset: val (5105 images) and train (4998 images). The dataset was released in 2014 by the Stanford University.
Here is a visualized example for randomly selected sample classes:
Explore #
PASCAL Context dataset has 10103 images. Click on one of the examples below or open "Explore" tool anytime you need to view dataset images with annotations. This tool has extended visualization capabilities like zoom, translation, objects table, custom filters and more. Hover the mouse over the images to hide or show annotations.
Class balance #
There are 459 annotation classes in the dataset. Find the general statistics and balances for every class in the table below. Click any row to preview images that have labels of the selected class. Sort by column to find the most rare or prevalent classes.
Class ㅤ | Images ㅤ | Objects ㅤ | Count on image average | Area on image average |
---|---|---|---|---|
person➔ mask | 3919 | 3919 | 1 | 20.65% |
unknown➔ mask | 3734 | 3734 | 1 | 9% |
tree➔ mask | 3232 | 3232 | 1 | 18.03% |
sky➔ mask | 3231 | 3231 | 1 | 26.77% |
building➔ mask | 3064 | 3064 | 1 | 20.23% |
ground➔ mask | 2987 | 2987 | 1 | 20.98% |
wall➔ mask | 2794 | 2794 | 1 | 22.24% |
grass➔ mask | 2742 | 2742 | 1 | 23.28% |
floor➔ mask | 1895 | 1895 | 1 | 16.07% |
table➔ mask | 1566 | 1566 | 1 | 9.77% |
Images #
Explore every single image in the dataset with respect to the number of annotations of each class it has. Click a row to preview selected image. Sort by any column to find anomalies and edge cases. Use horizontal scroll if the table has many columns for a large number of classes in the dataset.
Object distribution #
Interactive heatmap chart for every class with object distribution shows how many images are in the dataset with a certain number of objects of a specific class. Users can click cell and see the list of all corresponding images.
Class sizes #
The table below gives various size properties of objects for every class. Click a row to see the image with annotations of the selected class. Sort columns to find classes with the smallest or largest objects or understand the size differences between classes.
Class | Object count | Avg area | Max area | Min area | Min height | Min height | Max height | Max height | Avg height | Avg height | Min width | Min width | Max width | Max width |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
person mask | 3919 | 20.65% | 95.72% | 0% | 1px | 0.2% | 500px | 100% | 246px | 62.67% | 1px | 0.2% | 500px | 100% |
unknown mask | 3734 | 9% | 95.84% | 0% | 1px | 0.2% | 500px | 100% | 173px | 44.11% | 1px | 0.2% | 500px | 100% |
tree mask | 3232 | 18.03% | 99.86% | 0% | 1px | 0.3% | 500px | 100% | 173px | 44.92% | 1px | 0.2% | 500px | 100% |
sky mask | 3231 | 26.77% | 99.93% | 0% | 1px | 0.27% | 500px | 100% | 158px | 42.27% | 1px | 0.2% | 500px | 100% |
building mask | 3064 | 20.23% | 99.73% | 0% | 1px | 0.3% | 500px | 100% | 189px | 49.13% | 1px | 0.2% | 500px | 100% |
ground mask | 2987 | 20.98% | 97.08% | 0% | 1px | 0.3% | 500px | 100% | 176px | 46.15% | 1px | 0.2% | 500px | 100% |
wall mask | 2794 | 22.24% | 92.97% | 0.01% | 4px | 1.07% | 500px | 100% | 260px | 66.14% | 4px | 0.8% | 500px | 100% |
grass mask | 2742 | 23.28% | 99.71% | 0% | 1px | 0.27% | 500px | 100% | 184px | 48.53% | 1px | 0.2% | 500px | 100% |
floor mask | 1895 | 16.07% | 91.12% | 0.01% | 5px | 1% | 500px | 100% | 174px | 44.56% | 5px | 1% | 500px | 100% |
table mask | 1566 | 9.77% | 66.76% | 0.04% | 6px | 1.2% | 500px | 100% | 157px | 40.42% | 10px | 2% | 500px | 100% |
Spatial Heatmap #
The heatmaps below give the spatial distributions of all objects for every class. These visualizations provide insights into the most probable and rare object locations on the image. It helps analyze objects' placements in a dataset.
Objects #
Table contains all 65937 objects. Click a row to preview an image with annotations, and use search or pagination to navigate. Sort columns to find outliers in the dataset.
Object ID ㅤ | Class ㅤ | Image name click row to open | Image size height x width | Height ㅤ | Height ㅤ | Width ㅤ | Width ㅤ | Area ㅤ |
---|---|---|---|---|---|---|---|---|
1➔ | bird mask | 2009_003685.jpeg | 500 x 369 | 424px | 84.8% | 226px | 61.25% | 30.13% |
2➔ | grass mask | 2009_003685.jpeg | 500 x 369 | 144px | 28.8% | 117px | 31.71% | 7.64% |
3➔ | ground mask | 2009_003685.jpeg | 500 x 369 | 187px | 37.4% | 369px | 100% | 17.57% |
4➔ | sky mask | 2009_003685.jpeg | 500 x 369 | 197px | 39.4% | 369px | 100% | 25.59% |
5➔ | tree mask | 2009_003685.jpeg | 500 x 369 | 195px | 39% | 369px | 100% | 11.3% |
6➔ | wood mask | 2009_003685.jpeg | 500 x 369 | 118px | 23.6% | 139px | 37.67% | 7.77% |
7➔ | disc case mask | 2008_005006.jpeg | 500 x 497 | 86px | 17.2% | 252px | 50.7% | 7.02% |
8➔ | floor mask | 2008_005006.jpeg | 500 x 497 | 73px | 14.6% | 90px | 18.11% | 2.39% |
9➔ | shelves mask | 2008_005006.jpeg | 500 x 497 | 500px | 100% | 426px | 85.71% | 37.48% |
10➔ | table mask | 2008_005006.jpeg | 500 x 497 | 194px | 38.8% | 261px | 52.52% | 16.97% |
License #
The VOC2010 data includes images obtained from the “flickr” website. Use of these images must respect the corresponding terms of use:
For the purposes of the challenge, the identity of the images in the database, e.g. source and name of owner, has been obscured. Details of the contributor of each image can be found in the annotation to be included in the final release of the data, after completion of the challenge. Any queries about the use or ownership of the data should be addressed to the organizers.
Citation #
If you make use of the PASCAL context data, please cite the following reference:
@InProceedings{mottaghi_cvpr14,
author = {Roozbeh Mottaghi and Xianjie Chen and Xiaobai Liu and Nam-Gyu Cho and Seong-Whan Lee and Sanja Fidler and Raquel Urtasun and Alan Yuille},
title = {The Role of Context for Object Detection and Semantic Segmentation in the Wild},
booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2014},
}
If you are happy with Dataset Ninja and use provided visualizations and tools in your work, please cite us:
@misc{ visualization-tools-for-pascal-context-dataset,
title = { Visualization Tools for PASCAL Context Dataset },
type = { Computer Vision Tools },
author = { Dataset Ninja },
howpublished = { \url{ https://datasetninja.com/pascal-context } },
url = { https://datasetninja.com/pascal-context },
journal = { Dataset Ninja },
publisher = { Dataset Ninja },
year = { 2024 },
month = { oct },
note = { visited on 2024-10-15 },
}
Download #
Dataset PASCAL Context can be downloaded in Supervisely format:
As an alternative, it can be downloaded with dataset-tools package:
pip install --upgrade dataset-tools
… using following python code:
import dataset_tools as dtools
dtools.download(dataset='PASCAL Context', dst_dir='~/dataset-ninja/')
Make sure not to overlook the python code example available on the Supervisely Developer Portal. It will give you a clear idea of how to effortlessly work with the downloaded dataset.
The data in original format can be downloaded here:
Disclaimer #
Our gal from the legal dep told us we need to post this:
Dataset Ninja provides visualizations and statistics for some datasets that can be found online and can be downloaded by general audience. Dataset Ninja is not a dataset hosting platform and can only be used for informational purposes. The platform does not claim any rights for the original content, including images, videos, annotations and descriptions. Joint publishing is prohibited.
You take full responsibility when you use datasets presented at Dataset Ninja, as well as other information, including visualizations and statistics we provide. You are in charge of compliance with any dataset license and all other permissions. You are required to navigate datasets homepage and make sure that you can use it. In case of any questions, get in touch with us at hello@datasetninja.com.