SODA: A large-scale Small Object Detection dAtaset

SODA is a large-scale benckmark for Small Object Detection, including SODA-D and SODA-A, which concentrate on Driving and Aerial scenarios respectively.

Update

• [20220726] Our Homepage for SODA benchmark opens!
• [20220727] We add the visualization of images in SODA.
• [20220728] The quantitative results of several mainstream methods on SODA test-set are reported!
• [20220728] Our paper of SODA dataset is submitted to arXiv.
• [20221225] SODA Benchmark is officially available, please refer to the Download below.
• [20230102] Benchmark codes of SODA dataset are available now, please refer to the Codes part.
• [20230625] Our paper has been accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), in which the final version and Appendix files are available.
• [20231004] One paper based on the SODA benchmark has been accepted by ICCV 2023, and the code please refer to the Project.

SODA-D

SODA-D contains 24828 well-chosen and high-quality images under driving scenarios, on which 278433 instances of 9 categories with horizontal bounding boxes were annotated. Some example images are shown below.

SODA-A

SODA-A comprises 2513 high-resolution images of aerial scenes, which has 872069 instances annotated with oriented rectangle box annotations over 9 classes. An example image of SODA-A is shown next.

Results

We exhibit the quantitative experiment resullts of several representative methods on SODA test-set as follow.

• SODA-D

• SODA-A

Download

We provide the downloading of our paper and datasets.

• Paper: Towards Large-Scale Small Object Detection: Survey and Benchmarks
• Datasets:
  • SODA-D: OneDrvie; BaiduNetDisk
  • SODA-A: OneDrvie; BaiduNetDisk

Codes

The official codes of our benchmark, which mainly includes data preparation and evaluation, are released below.

• SODA-D Benchmark: SODA-mmdetection
• SODA-A Benchmark: SODA-mmrotate

Support or Contact

If you have any problems about our SODA benchmark, please be free to contact us at shaunyuan@mail.nwpu.edu.cn.