{"id":13738287,"url":"https://github.com/fcakyon/small-object-detection-benchmark","last_synced_at":"2025-04-04T19:10:36.918Z","repository":{"id":37822310,"uuid":"439981746","full_name":"fcakyon/small-object-detection-benchmark","owner":"fcakyon","description":"icip2022 paper: sahi benchmark on visdrone and xview datasets using fcos, vfnet and tood detectors","archived":false,"fork":false,"pushed_at":"2025-01-17T00:46:56.000Z","size":112,"stargazers_count":171,"open_issues_count":0,"forks_count":25,"subscribers_count":2,"default_branch":"main","last_synced_at":"2025-03-28T18:12:47.385Z","etag":null,"topics":["benchmark","coco","fcos","icip2022","mmdetection","object-detection","pytorch","sahi","small-object-detection","tood","vfnet","visdrone","xview"],"latest_commit_sha":null,"homepage":"https://ieeexplore.ieee.org/document/9897990","language":"Python","has_issues":false,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/fcakyon.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":"CITATION.cff","codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2021-12-19T22:27:14.000Z","updated_at":"2025-03-10T14:21:50.000Z","dependencies_parsed_at":"2025-01-16T14:15:10.943Z","dependency_job_id":"0d8efa91-5adc-43ac-9de5-0c98b08bce6d","html_url":"https://github.com/fcakyon/small-object-detection-benchmark","commit_stats":{"total_commits":56,"total_committers":3,"mean_commits":"18.666666666666668","dds":0.4821428571428571,"last_synced_commit":"cf7631dd0db1821b60422cfc699610dfb134632a"},"previous_names":[],"tags_count":2,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fcakyon%2Fsmall-object-detection-benchmark","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fcakyon%2Fsmall-object-detection-benchmark/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fcakyon%2Fsmall-object-detection-benchmark/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/fcakyon%2Fsmall-object-detection-benchmark/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/fcakyon","download_url":"https://codeload.github.com/fcakyon/small-object-detection-benchmark/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247234921,"owners_count":20905854,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["benchmark","coco","fcos","icip2022","mmdetection","object-detection","pytorch","sahi","small-object-detection","tood","vfnet","visdrone","xview"],"created_at":"2024-08-03T03:02:17.226Z","updated_at":"2025-04-04T19:10:36.892Z","avatar_url":"https://github.com/fcakyon.png","language":"Python","funding_links":[],"categories":["Python"],"sub_categories":[],"readme":"# small-object-detection-benchmark\n\n\u003ca href=\"https://ieeexplore.ieee.org/document/9897990\"\u003e\u003cimg src=\"https://img.shields.io/badge/DOI-10.1109%2FICIP46576.2022.9897990-orange.svg\" alt=\"ci\"\u003e\n\u003ca href=\"https://twitter.com/fcakyon\"\u003e\u003cimg src=\"https://img.shields.io/badge/twitter-fcakyon_-blue?logo=twitter\u0026style=flat\" alt=\"fcakyon twitter\"\u003e\u003c/a\u003e\n\n🔥 our paper has been presented in ICIP 2022 Bordeaux, France (16-19 October 2022)\n\n[📜 List of publications that cite this work (currently 300+)](https://scholar.google.com/scholar?hl=en\u0026as_sdt=2005\u0026sciodt=0,5\u0026cites=14065474760484865747\u0026scipsc=\u0026q=\u0026scisbd=1)\n\n## summary\n\nsmall-object-detection benchmark on visdrone and xview datasets using [fcos](https://arxiv.org/abs/1904.01355), [vfnet](https://arxiv.org/abs/2008.13367) and [tood](https://arxiv.org/abs/2108.07755) detectors\n\nrefer to [Slicing Aided Hyper Inference and Fine-tuning for Small Object Detection](https://ieeexplore.ieee.org/document/9897990) for full technical analysis\n\n## citation\n\nIf you use any file/result from this repo in your work, please cite it as:\n\n```\n@article{akyon2022sahi,\n  title={Slicing Aided Hyper Inference and Fine-tuning for Small Object Detection},\n  author={Akyon, Fatih Cagatay and Altinuc, Sinan Onur and Temizel, Alptekin},\n  journal={2022 IEEE International Conference on Image Processing (ICIP)},\n  doi={10.1109/ICIP46576.2022.9897990},\n  pages={966-970},\n  year={2022}\n}\n```\n\n## visdrone results\n\nrefer to table 1 in [Slicing Aided Hyper Inference and Fine-tuning for Small Object Detection](https://ieeexplore.ieee.org/document/9897990) for more detail on visdrone results\n\n[fcos_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_fi_visdrone_results.zip\n[fcos_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sahi_po_visdrone_results.zip\n[fcos_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sahi_fi_po_visdrone_results.zip\n[fcos_sf_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_po_visdrone_results.zip\n[fcos_sf_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_fi_po_visdrone_results.zip\n\n[vfnet_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_fi_visdrone_results.zip\n[vfnet_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sahi_po_visdrone_results.zip\n[vfnet_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sahi_fi_po_visdrone_results.zip\n[vfnet_sf_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_po_visdrone_results.zip\n[vfnet_sf_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_fi_po_visdrone_results.zip\n\n[tood_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_fi_visdrone_results.zip\n[tood_sahi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sahi_visdrone_results.zip\n[tood_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sahi_po_visdrone_results.zip\n[tood_sahi_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sahi_fi_visdrone_results.zip\n[tood_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sahi_fi_po_visdrone_results.zip\n\n[tood_sf_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_fi_visdrone_results.zip\n[tood_sf_sahi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_visdrone_results.zip\n[tood_sf_sahi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_po_visdrone_results.zip\n[tood_sf_sahi_fi_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_fi_visdrone_results.zip\n[tood_sf_sahi_fi_po_visdrone_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_fi_po_visdrone_results.zip\n\n[tood_sf_visdrone_checkpoint_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.2/tood_sf_visdrone.pth\n[fcos_sf_visdrone_checkpoint_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.2/fcos_sf_visdrone.pth\n\n[my_twitter_url]: https://twitter.com/fcakyon\n\n|setup |AP\u003csub\u003e50\u003c/sub\u003e |AP\u003csub\u003e50\u003c/sub\u003es |AP\u003csub\u003e50\u003c/sub\u003em |AP\u003csub\u003e50\u003c/sub\u003el | results | checkpoints |\n|--- |--- |--- |--- |--- |--- |--- |\n|FCOS+FI |25.8 |14.2 |39.6 |45.1 | [download][fcos_fi_visdrone_results_url] | [request][my_twitter_url] |\n|FCOS+SAHI+PO |29.0 |18.9 |41.5 |46.4 | [download][fcos_sahi_po_visdrone_results_url] | [request][my_twitter_url] |\n|FCOS+SAHI+FI+PO |31.0 |19.8 |44.6 |49.0 | [download][fcos_sahi_fi_po_visdrone_results_url] | [request][my_twitter_url] |\n|FCOS+SF+SAHI+PO |38.1 |25.7 |54.8 |56.9 | [download][fcos_sf_sahi_po_visdrone_results_url] | [download][fcos_sf_visdrone_checkpoint_url] |\n|FCOS+SF+SAHI+FI+PO |38.5 |25.9 |55.4 |59.8 | [download][fcos_sf_sahi_fi_po_visdrone_results_url] | [download][fcos_sf_visdrone_checkpoint_url] |\n|--- |--- |--- |--- |--- |--- |--- |\n|VFNet+FI |28.8 |16.8 |44.0 |47.5 | [download][vfnet_fi_visdrone_results_url] | [request][my_twitter_url] |\n|VFNet+SAHI+PO |32.0 |21.4 |45.8 |45.5 | [download][vfnet_sahi_po_visdrone_results_url] | [request][my_twitter_url] |\n|VFNet+SAHI+FI+PO |33.9 |22.4 |49.1 |49.4 | [download][vfnet_sahi_fi_po_visdrone_results_url] | [request][my_twitter_url] |\n|VFNet+SF+SAHI+PO |41.9 |29.7 |58.8 |60.6 | [download][vfnet_sf_sahi_po_visdrone_results_url] | [request][my_twitter_url] |\n|VFNet+SF+SAHI+FI+PO |42.2 |29.6 |59.2 |63.3 | [download][vfnet_sf_sahi_fi_po_visdrone_results_url] | [request][my_twitter_url] |\n|--- |--- |--- |--- |--- |--- |--- |\n|TOOD+FI |29.4 |18.1 |44.1 |50.0 | [download][tood_fi_visdrone_results_url] | [request][my_twitter_url] |\n|TOOD+SAHI |31.9 |22.6 |44.0 |45.2 | [download][tood_sahi_visdrone_results_url] | [request][my_twitter_url] |\n|TOOD+SAHI+PO |32.5 |22.8 |45.2 |43.6 | [download][tood_sahi_po_visdrone_results_url] | [request][my_twitter_url] |\n|TOOD+SAHI+FI |34.6 |23.8 |48.5 |53.1 | [download][tood_sahi_fi_visdrone_results_url] | [request][my_twitter_url] |\n|TOOD+SAHI+FI+PO |34.7 |23.8 |48.9 |50.3| [download][tood_sahi_fi_po_visdrone_results_url] | [request][my_twitter_url] |\n|TOOD+SF+FI |36.8 |24.4 |53.8 |66.4 | [download][tood_sf_fi_visdrone_results_url] | [download][tood_sf_visdrone_checkpoint_url] |\n|TOOD+SF+SAHI |42.5 |31.6 |58.0 |61.1 | [download][tood_sf_sahi_visdrone_results_url] | [download][tood_sf_visdrone_checkpoint_url] |\n|TOOD+SF+SAHI+PO |43.1 |31.7 |59.0 |60.2 | [download][tood_sf_sahi_po_visdrone_results_url] | [download][tood_sf_visdrone_checkpoint_url] |\n|TOOD+SF+SAHI+FI |43.4 |31.7 |59.6 |65.6 | [download][tood_sf_sahi_fi_visdrone_results_url] | [download][tood_sf_visdrone_checkpoint_url] |\n|TOOD+SF+SAHI+FI+PO |43.5 |31.7 |59.8 |65.4 | [download][tood_sf_sahi_fi_po_visdrone_results_url] | [download][tood_sf_visdrone_checkpoint_url] |\n\n## xview results\n\nrefer to table 2 in [Slicing Aided Hyper Inference and Fine-tuning for Small Object Detection](https://ieeexplore.ieee.org/document/9897990) for more detail on xview results\n\n[fcos_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_fi_xview_results.zip\n[fcos_sf_sahi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_xview_results.zip\n[fcos_sf_sahi_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_fi_xview_results.zip\n[fcos_sf_sahi_fi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_fi_op_xview_results.zip\n[fcos_sf_sahi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/fcos_sf_sahi_op_xview_results.zip\n\n[vfnet_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_fi_xview_results.zip\n[vfnet_sf_sahi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_xview_results.zip\n[vfnet_sf_sahi_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_fi_xview_results.zip\n[vfnet_sf_sahi_fi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_fi_op_xview_results.zip\n[vfnet_sf_sahi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/vfnet_sf_sahi_op_xview_results.zip\n\n[tood_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_fi_xview_results.zip\n[tood_sf_sahi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_xview_results.zip\n[tood_sf_sahi_fi_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_fi_xview_results.zip\n[tood_sf_sahi_fi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_fi_op_xview_results.zip\n[tood_sf_sahi_po_xview_results_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.1/tood_sf_sahi_op_xview_results.zip\n\n[fcos_sf_xview_checkpoint_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.2/fcos_sf_xview.pth\n[vfnet_sf_xview_checkpoint_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.2/vfnet_sf_xview.pth\n[tood_sf_xview_checkpoint_url]: https://github.com/fcakyon/sahi-benchmark/releases/download/v0.0.2/tood_sf_xview.pth\n\n|setup |AP\u003csub\u003e50\u003c/sub\u003e |AP\u003csub\u003e50\u003c/sub\u003es |AP\u003csub\u003e50\u003c/sub\u003em |AP\u003csub\u003e50\u003c/sub\u003el | results | checkpoints |\n|--- |--- |--- |--- |--- |--- |--- |\n|FCOS+FI |2.20 |0.10 |1.80 |7.30 | [download][fcos_fi_xview_results_url] | [request][my_twitter_url] |\n|FCOS+SF+SAHI |15.8 |11.9 |18.4 |11.0 | [download][fcos_sf_sahi_xview_results_url] | [download][fcos_sf_xview_checkpoint_url] |\n|FCOS+SF+SAHI+PO |17.1 |12.2 |20.2 |12.8 | [download][fcos_sf_sahi_po_xview_results_url] | [download][fcos_sf_xview_checkpoint_url] |\n|FCOS+SF+SAHI+FI |15.7 |11.9 |18.4 |14.3 | [download][fcos_sf_sahi_fi_xview_results_url] | [download][fcos_sf_xview_checkpoint_url] |\n|FCOS+SF+SAHI+FI+PO |17.0 |12.2 |20.2 |15.8 | [download][fcos_sf_sahi_fi_po_xview_results_url] | [download][fcos_sf_xview_checkpoint_url] |\n|--- |--- |--- |--- |--- |--- |--- |\n|VFNet+FI |2.10 |0.50 |1.80 |6.80 | [download][vfnet_fi_xview_results_url] | [request][my_twitter_url] |\n|VFNet+SF+SAHI | 16.0 |11.9 |17.6 |13.1 | [download][vfnet_sf_sahi_xview_results_url] | [download][vfnet_sf_xview_checkpoint_url] |\n|VFNet+SF+SAHI+PO |17.7| 13.7 |19.7 |15.4 | [download][vfnet_sf_sahi_po_xview_results_url] | [download][vfnet_sf_xview_checkpoint_url] |\n|VFNet+SF+SAHI+FI |15.8 |11.9 |17.5 |15.2 | [download][vfnet_sf_sahi_fi_xview_results_url] | [download][vfnet_sf_xview_checkpoint_url] |\n|VFNet+SF+SAHI+FI+PO |17.5 |13.7 |19.6 |17.6 | [download][vfnet_sf_sahi_fi_po_xview_results_url] | [download][vfnet_sf_xview_checkpoint_url] |\n|--- |--- |--- |--- |--- |--- |--- |\n|TOOD+FI |2.10 |0.10 |2.00 |5.20 | [download][tood_fi_xview_results_url] | [request][my_twitter_url] |\n|TOOD+SF+SAHI |19.4 |14.6 |22.5 |14.2 | [download][tood_sf_sahi_xview_results_url] | [download][tood_sf_xview_checkpoint_url] |\n|TOOD+SF+SAHI+PO |20.6 |14.9 |23.6 |17.0 | [download][tood_sf_sahi_po_xview_results_url] | [download][tood_sf_xview_checkpoint_url] |\n|TOOD+SF+SAHI+FI |19.2 |14.6 |22.3 |14.7 | [download][tood_sf_sahi_fi_xview_results_url] | [download][tood_sf_xview_checkpoint_url] |\n|TOOD+SF+SAHI+FI+PO |20.4 |14.9 |23.5 |17.6 | [download][tood_sf_sahi_fi_po_xview_results_url] | [download][tood_sf_xview_checkpoint_url] |\n\n## env setup\n\ninstall pytorch:\n\n```bash\nconda install pytorch=1.10.0 torchvision=0.11.1 cudatoolkit=11.3 -c pytorch\n```\n\ninstall other requirements:\n\n```bash\npip install -r requirements.txt\n```\n\n## evaluation\n\n- download desired checkpoint from the urls in readme.\n\n- download xivew or visdrone dataset and convert to COCO format.\n\n- set `MODEL_PATH`, `MODEL_CONFIG_PATH`, `EVAL_IMAGES_FOLDER_DIR`, `EVAL_DATASET_JSON_PATH`, `INFERENCE_SETTING` in [predict_evaluate_analyse script](eval_tools/predict_evaluate_analyse.py) then run the script.\n\n## roadmap\n\n- [x] add train test split support for xview to coco converter\n- [x] add mmdet config files (fcos, vfnet and tood) for xview training (9 train experiments)\n- [x] add mmdet config files (fcos, vfnet and tood) for visdrone training (9 train experiments)\n- [x] add coco result.json files, classwise coco eval results error analysis plots for all xview experiments\n- [x] add coco result.json files, classwise coco eval results error analysis plots for all visdrone experiments\n- [X] add .py scripts for inference + evaluation + error analysis using `sahi`\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ffcakyon%2Fsmall-object-detection-benchmark","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ffcakyon%2Fsmall-object-detection-benchmark","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ffcakyon%2Fsmall-object-detection-benchmark/lists"}