AFO - Dataset Ninja

Introduction #

Jan Gąsienica-Józkowy, Mateusz Knapik, Boguslaw Cyganek

The authors of the AFO: Aerial Dataset of Floating Objects present a comprehensive study focusing on leveraging deep learning architectures for marine search and rescue operations involving object detection. This dataset consists of 3647 images, each manually annotated with nearly 40,000 labeled objects.

Highlighting the significance of the study, the authors discuss the challenges associated with search and rescue operations, emphasizing the difficulties in localizing missing individuals and objects. They emphasize the potential benefits of incorporating unmanned aerial vehicles, artificial intelligence, and computer vision technologies to enhance rescue missions. While these technologies offer promising assistance, the authors underscore the scarcity of object detection systems based on neural networks in SAR applications, partly due to the lack of sufficiently large training datasets.

To address this limitation, the authors present the AFO dataset, which is meticulously curated for the problem of detecting small objects in marine environments. The dataset is divided into three subsets: train, test, and validation. The authors also elaborate on the considerations for data selection, including altitude and angle restrictions for video recordings, as well as the inclusion of various environmental and weather conditions in the dataset.

Each class here belongs to one of three dataset versions (depending on the goal of the CV task): 6categories version (classes human, wind/sup-board, kayak, boat, bouy, sailboat) to check how accurately detectors can detect humans vs. other floating objects (a large data imbalance ); 2categories version (classes small_obj, large_obj) to the task of searching for missing people (small objects) and boats (large objects) (slightly more balanced ); 1category version (class object) which marks all bounding boxes as one class to reflect the fact that usually during search and rescue operations at the sea, finding any object of a human origin can be significant.

Expand

Homepage

Research Paper

Summary #

AFO: Aerial Dataset of Floating Objects is a dataset for an object detection task. It is used in the search and rescue (SAR) industry.

The dataset consists of 3640 images with 119973 labeled objects belonging to 9 different classes including object, small_obj, human, and other: large_obj, wind/sup-board, kayak, boat, bouy, and sailboat.

Images in the AFO dataset have bounding box annotations. There are 728 (20% of the total) unlabeled images (i.e. without annotations). There are 3 splits in the dataset: train (2787 images), test (514 images), and validation (339 images). The dataset was released in 2021 by the AGH University of Science and Technology, Poland.

Explore #

AFO dataset has 3640 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.

👀

Have a look at 3640 images

View images along with annotations and tags, search and filter by various parameters

Class balance #

There are 9 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.

Rows 1-9 of 9

Class ㅤ	Images ㅤ	Objects ㅤ	Count on image average	Area on image average
object➔ rectangle	2912	39991	13.73	1.89%
small_obj➔ rectangle	2124	33761	15.9	1.42%
human➔ rectangle	2107	33174	15.74	1.43%
large_obj➔ rectangle	1931	6230	3.23	1.45%
wind/sup-board➔ rectangle	1025	3922	3.83	1.35%
kayak➔ rectangle	546	1446	2.65	0.87%
boat➔ rectangle	339	702	2.07	1.46%
bouy➔ rectangle	177	587	3.32	0.06%
sailboat➔ rectangle	131	160	1.22	3.56%

Co-occurrence matrix #

Co-occurrence matrix is an extremely valuable tool that shows you the images for every pair of classes: how many images have objects of both classes at the same time. If you click any cell, you will see those images. We added the tooltip with an explanation for every cell for your convenience, just hover the mouse over a cell to preview the description.

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.

Rows 1-9 of 9

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
object rectangle	39991	0.15%	21.31%	0%	6px	0.28%	1164px	80.83%	92px	4.48%	11px	0.42%	1256px	49.06%
small_obj rectangle	33761	0.09%	1.1%	0%	10px	0.69%	308px	15.28%	80px	3.8%	11px	0.42%	293px	9.53%
human rectangle	33174	0.09%	1.1%	0.01%	16px	0.79%	308px	15.28%	81px	3.84%	14px	0.42%	293px	9.53%
large_obj rectangle	6230	0.46%	21.31%	0%	6px	0.28%	1164px	80.83%	155px	8.15%	19px	0.74%	1256px	49.06%
wind/sup-board rectangle	3922	0.36%	0.92%	0%	6px	0.28%	345px	15.97%	164px	7.61%	32px	0.83%	382px	9.95%
kayak rectangle	1446	0.33%	1.2%	0.02%	18px	1.18%	254px	16.6%	119px	7.8%	35px	1.29%	222px	8.16%
boat rectangle	702	0.72%	16.19%	0.02%	22px	1.53%	657px	34.38%	130px	8.1%	19px	0.74%	1256px	49.06%
bouy rectangle	587	0.02%	0.07%	0%	10px	0.69%	75px	3.47%	37px	1.75%	11px	0.43%	86px	2.24%
sailboat rectangle	160	2.96%	21.31%	0.14%	130px	9.03%	1164px	80.83%	360px	24.73%	40px	1.56%	675px	26.37%

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 119973 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.

Rows 1-10 of 119973

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➔	object rectangle	r3_255.jpg	1530 x 2720	107px	6.99%	100px	3.68%	0.26%
2➔	object rectangle	r3_255.jpg	1530 x 2720	105px	6.86%	76px	2.79%	0.19%
3➔	object rectangle	r3_255.jpg	1530 x 2720	84px	5.49%	54px	1.99%	0.11%
4➔	large_obj rectangle	r3_255.jpg	1530 x 2720	107px	6.99%	100px	3.68%	0.26%
5➔	large_obj rectangle	r3_255.jpg	1530 x 2720	105px	6.86%	76px	2.79%	0.19%
6➔	large_obj rectangle	r3_255.jpg	1530 x 2720	84px	5.49%	54px	1.99%	0.11%
7➔	kayak rectangle	r3_255.jpg	1530 x 2720	107px	6.99%	100px	3.68%	0.26%
8➔	kayak rectangle	r3_255.jpg	1530 x 2720	105px	6.86%	76px	2.79%	0.19%
9➔	kayak rectangle	r3_255.jpg	1530 x 2720	84px	5.49%	54px	1.99%	0.11%
10➔	object rectangle	i3_137.jpg	1440 x 2560	584px	40.56%	357px	13.95%	5.66%

License #

AFO: Aerial Dataset of Floating Objects is under CC BY-NC-SA 3.0 IGO license.

Source

Citation #

If you make use of the AFO data, please cite the following reference:

@article{article,
    author = {Gąsienica-Józkowy, Jan and Knapik, Mateusz and Cyganek, Boguslaw},
    year = {2021},
    month = {01},
    pages = {1-15},
    title = {An ensemble deep learning method with optimized weights for drone-based water rescue and surveillance},
    journal = {Integrated Computer-Aided Engineering},
    doi = {10.3233/ICA-210649}
}

Source

If you are happy with Dataset Ninja and use provided visualizations and tools in your work, please cite us:

@misc{ visualization-tools-for-afo-dataset,
  title = { Visualization Tools for AFO Dataset },
  type = { Computer Vision Tools },
  author = { Dataset Ninja },
  howpublished = { \url{ https://datasetninja.com/afo } },
  url = { https://datasetninja.com/afo },
  journal = { Dataset Ninja },
  publisher = { Dataset Ninja },
  year = { 2024 },
  month = { jul },
  note = { visited on 2024-07-27 },
}

Download #

Dataset AFO 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='AFO', 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.