{"id":13443506,"url":"https://github.com/zhangzhengde0225/CDNet","last_synced_at":"2025-03-20T16:31:50.154Z","repository":{"id":37362058,"uuid":"302567880","full_name":"zhangzhengde0225/CDNet","owner":"zhangzhengde0225","description":"The tutorials, datasets and source codes of the crosswalk detection (zebra crossing detection) network, which is robust in real scenes and real-time in Jetson nano. cross. detect. pedestrian.","archived":false,"fork":false,"pushed_at":"2022-04-01T16:56:14.000Z","size":15734,"stargazers_count":135,"open_issues_count":9,"forks_count":28,"subscribers_count":2,"default_branch":"master","last_synced_at":"2024-08-01T03:43:51.363Z","etag":null,"topics":["crosswalk","crosswalk-detection","deep-learning","jetson-nano","nvidia","object-detection","zebra","zebra-crossing"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/zhangzhengde0225.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2020-10-09T07:35:48.000Z","updated_at":"2024-07-31T08:06:06.000Z","dependencies_parsed_at":"2022-08-03T04:15:48.072Z","dependency_job_id":null,"html_url":"https://github.com/zhangzhengde0225/CDNet","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhangzhengde0225%2FCDNet","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhangzhengde0225%2FCDNet/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhangzhengde0225%2FCDNet/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/zhangzhengde0225%2FCDNet/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/zhangzhengde0225","download_url":"https://codeload.github.com/zhangzhengde0225/CDNet/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":221780032,"owners_count":16879040,"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":["crosswalk","crosswalk-detection","deep-learning","jetson-nano","nvidia","object-detection","zebra","zebra-crossing"],"created_at":"2024-07-31T03:02:02.380Z","updated_at":"2024-10-28T04:31:17.693Z","avatar_url":"https://github.com/zhangzhengde0225.png","language":"Python","funding_links":[],"categories":["Python","Object Detection Applications"],"sub_categories":[],"readme":"[![Stars](https://img.shields.io/github/stars/zhangzhengde0225/CDNet)](\nhttps://github.com/zhangzhengde0225/CDNet)\n[![Open issue](https://img.shields.io/github/issues/zhangzhengde0225/CDNet)](\nhttps://github.com/zhangzhengde0225/CDNet/issues)\n[![Datasets](https://img.shields.io/static/v1?label=Download\u0026message=datasets\u0026color=green)](\nhttps://github.com/zhangzhengde0225/CDNet/blob/master/docs/DATASETS.md)\n[![Datasets](https://img.shields.io/static/v1?label=Download\u0026message=source_code\u0026color=orange)](\nhttps://github.com/zhangzhengde0225/CDNet/archive/refs/heads/master.zip)\n\n#### English | [简体中文](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/README_zh_cn.md)\n\nPlease **star this project** in the upper right corner and **cite this paper** blow \nif this project helps you. \n\n# CDNet\n\nThis repository is the codes, datasets and tutorials for the paper \n\"CDNet: a real-time and robust crosswalk detection network on Jetson nano based on YOLOv5\".\n\nCDNet (Crosswalk Detection Network) is a specific implementation of crosswalk (zebra crossing) detection and vehicle crossing behavior analysis under the vision of vehicle-mounted camera. \n\n![GA](https://zhangzhengde0225.github.io/images/CDNet_GA.jpg)\n\nFig.1 Graphical abstract.\n\n# Highlights\n+ A crosswalk detection and vehicle crossing behavior detection network is realized.\n+ The accuracy and speed exceed YOLOv5 in the specific task.\n+ High robustness in real complex scenarios such as in cloudy, sunny, rainy and at night is achieved.\n+ Real-time detection (33.1 FPS) is implemented on Jetson nano edge-computing device.\n+The datasets, tutorials and source codes are available on GitHub.\n  \n# Contribution\n\n+ SENet (Squeeze-and-Excitation Network), F1 score up, speed slightly down\n+ NST (Negative Samples Training), F1 score up, speed invariant\n+ ROI (Region Of Interest), F1 score down, speed up\n+ SSVM (Slide receptive field Short-term Vectors Memory), transfer crosswalk detection task into vehicle crossing behavior task, \n  F1 score up, speed invariant\n+ SFA (Synthetic Fog Augment), dataset augment, adapt to foggy weather, F1 score up, speed invariant\n\n# Installation\nGet CDNet code and configure the environment, please check out [docs/INSTALL.md](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/INSTALL.md)\n\n# Model Zoo\nPlease check out [docs/MODEL_ZOO.md](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/MODEL_ZOO.md)\n\n# Datasets\nDownload trainsets and testsets, please check out [docs/DATASETS.md](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/DATASETS.md)\n\n# Quick Start\n## Train\n\nOnce you get the CDNet code, configure the environment and download the dataset, juse type:\n```\npython train.py --trainset_path \u003c/path/to/trainset/folder\u003e\n(such as: /home/xxx/datasets/train_data_yolov5_format) \n```\nThe training results and weights will be saved in runs/expxx/ directory.\n\nThe main optional arguments:\n```\n--device \"0, 1\"  # cpu or gpu id, \"0, 1\" means use two gpu to train.\n--img-size 640 \n--batch-size 32 \n--epochs 100 \n--not-use-SE  # use original YOLOv5 which not SE-module embedded if there is the flag\n```\n\n## Inference\nDetect the crosswalk image by image and analyze the vehicle crossing behavior. \n```\npython detect.py\n```\n\nThe main optional arguments:\n```\n--source example/images  # images dir\n--output example/output  # output dir\n--img-size 640  # inference model size\n--device \"0\"   # use cpu or gpu(gpu id)\n--plot-classes [\"crosswalk\"]  # plot classes\n--field-size 5  # the Slide receptive field size of SSVM \n--not-use-ROI  # not use roi for accelerate inference speed if there is the flag\n--not-use-SSVM  # not use ssvm method for analyse vehicle crossing behavior if there is the flag\n```\n\nFor more details, please refer to [docs/INSTALL.md](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/INSTALL.md) and [docs/DATASETS.md](https://github.com/zhangzhengde0225/CDNet/blob/master/docs/DATASETS.md).\n\n## Fogging Augment\nIf you want to augment datasets by synthetic fog algorithm, just run:\n```\npython fog_augment.py\n```\nFor more details, please view the source code in fog_augment.py and /scripts/synthetic_fog.py\n\n# Results\n\n![Results](https://zhangzhengde0225.github.io/images/CDNet_Results.jpg)\n\nFig.2 Performance compared to YOLOv5.\n\n**CDNet improves the score for 5.13 points and speed for 10.7 FPS on Jetson nano for detection size of 640 compared to YOLOv5.**\n\n**For detection size of 288, the improvements are 13.38 points and 13.1 FPS.**\n\n\n# Contributors\nCDNet is authored by Zhengde Zhang, Menglu Tan, Zhicai Lan, Haichun Liu, Ling Pei and Wenxian Yu.\n\nCurrently, it is maintained by Zhengde Zhang (drivener@163.com).\n\nPlease feel free to contact us if you have any question.\n\nThe Academic homepage of Zhengde Zhang: [zhangzhengde0225.github.io](https://zhangzhengde0225.github.io).\n\n# Acknowledgement\n\nThis work was supported by the National Natural\nScience Foundation of China [Grant Numbers: 61873163]. \n\nWe acknowledge the Center for High Performance Computing at\nShanghai Jiao Tong University for providing computing resources.\n\nWe are very grateful to the \n[yolov5](https://github.com/ultralytics/yolov5) project\nfor the benchmark detection algorithm.\n\nWe are very grateful to the \n[tensorrtx](https://github.com/wang-xinyu/tensorrtx) project\nfor the deployment techniques to the Jetson nano.\n\n# Links\nDetect Video Samples：[https://www.bilibili.com/video/BV1qf4y1B7BA](https://www.bilibili.com/video/BV1qf4y1B7BA)\n\nRead Full Text of This Paper：[https://rdcu.be/cHuc8](https://rdcu.be/cHuc8)\n\nDownload Full Text of this Paper：[https://doi.org/10.1007/s00521-022-07007-9](https://doi.org/10.1007/s00521-022-07007-9)\n\nProject Introduction on CSDN：[http://t.csdn.cn/Cf7c7](http://t.csdn.cn/Cf7c7)\n\nIf it is helps you, \nplease star this project in the upper right corner and cite this paper blow.\n\n# Citation\n```\n@article{CDNet,\nauthor={Zheng-De Zhang, Meng-Lu Tan, Zhi-Cai Lan, Hai-Chun Liu, Ling Pei and Wen-Xian Yu},\ntitle={CDNet: a real-time and robust crosswalk detection network on Jetson nano based on YOLOv5},\nJournal={Neural Computing and Applications}, \nYear={2022},\nDOI={10.1007/s00521-022-07007-9},\n}\n```\n\n# License\nCDNet is freely available for free non-commercial use, and may be redistributed under these conditions. For commercial queries, please drop an e-mail at drivener@163.com. We will send the detail agreement to you.\n\n\n\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzhangzhengde0225%2FCDNet","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fzhangzhengde0225%2FCDNet","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fzhangzhengde0225%2FCDNet/lists"}