{"id":31964285,"url":"https://github.com/hanouticelina/deformable-detr","last_synced_at":"2025-10-14T17:13:21.577Z","repository":{"id":41468326,"uuid":"330717980","full_name":"hanouticelina/deformable-DETR","owner":"hanouticelina","description":"Implementation of the paper : Deformable DETR: Deformable Transformers for End-to-End Object Detection (ICLR 2021)","archived":false,"fork":false,"pushed_at":"2021-03-16T15:20:27.000Z","size":25768,"stargazers_count":19,"open_issues_count":1,"forks_count":7,"subscribers_count":1,"default_branch":"main","last_synced_at":"2023-05-30T04:05:12.161Z","etag":null,"topics":["computer-vision","deformable-convnets","detr","object-detection","transformers"],"latest_commit_sha":null,"homepage":"","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/hanouticelina.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2021-01-18T16:03:05.000Z","updated_at":"2023-05-13T13:38:54.000Z","dependencies_parsed_at":"2022-08-26T15:22:32.129Z","dependency_job_id":null,"html_url":"https://github.com/hanouticelina/deformable-DETR","commit_stats":null,"previous_names":[],"tags_count":null,"template":null,"template_full_name":null,"purl":"pkg:github/hanouticelina/deformable-DETR","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hanouticelina%2Fdeformable-DETR","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hanouticelina%2Fdeformable-DETR/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hanouticelina%2Fdeformable-DETR/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hanouticelina%2Fdeformable-DETR/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/hanouticelina","download_url":"https://codeload.github.com/hanouticelina/deformable-DETR/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/hanouticelina%2Fdeformable-DETR/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":279020087,"owners_count":26086805,"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","status":"online","status_checked_at":"2025-10-14T02:00:06.444Z","response_time":60,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"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":["computer-vision","deformable-convnets","detr","object-detection","transformers"],"created_at":"2025-10-14T17:13:12.380Z","updated_at":"2025-10-14T17:13:21.571Z","avatar_url":"https://github.com/hanouticelina.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"\n\n\n# Deformable DETR: Deformable Transformers for End-to-End Object Detection\n\n**Hakim Chekirou, Celina Hanouti \u0026 Aymen Merrouche (Equal Contribution)**\n\n\n![Poster](poster.jpg)\n\nThis repository contains an implementation of the paper \"Deformable Transformers for End-to-End Object Detection\" : https://arxiv.org/pdf/2010.04159.pdf \n\nWe provide scratch implementation of the following modules : \n\n* `deformable_transformer.py`, `decoder.py`, `encoder.py`, `deformable_detr.py` and `MultiHeadAttention.py`.\n\nthe remaining modules are mainly copied from the original DETR implementation : https://github.com/facebookresearch/detr\n\n\n## Requirements\n\n`pip install -r requirements.txt`\n\n## Dataset\nThe `coco_extraction.py` script provides functions for creating an annotation file for coco with only the specified class indexes. It removes all images and bounding boxes not containning at least one of those classes. We kept only five randomly sampled classes from 91 available, `bear`, `bus`, `tie`, `toilet` and `vase`. The annotation files are `datasets/coco_light/coco_light_train.json` for the train set and `datasets/coco_light/coco_light_train.json` for the validation set. `Coco lignt` contains ~ 15K images in the train set and 656 images on the validation set.\n\n\n## Usage \n\n`git clone https://github.com/hanouticelina/deformable-DETR.git`\n\n`cd deformable-DETR`\n\n### Training\n\nThe command for training Deformable DETR is as following:\n\n`python main.py --enc_layers 3 --dec_layers 3 --batch_size 1`\n\nTraining convergence takes 72 GPU hours on a single GPU GeForce RTX 2080.\n\n### Evaluation\n\nTo evaluate Deformable DETR on a subset of COCO 2017 validation set with a single GPU run:\n\n`\u003cpath to config file\u003e --resume \u003cpath to pre-trained model\u003e --eval`\n\nPre-trained model can be found at : https://www.dropbox.com/s/vnkbfrui1ldwtah/checkpoint.pth?dl=0\n\n\n## References\nXizhou Zhu et al., Deformable Transformers for End-to-End Object Detection.\n\nNicolas Carion et al., End-to-End Object Detection with Transformers.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhanouticelina%2Fdeformable-detr","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fhanouticelina%2Fdeformable-detr","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fhanouticelina%2Fdeformable-detr/lists"}