文件列表:

config.py (2037, 2021-01-06)
dataloader/ (0, 2021-01-06)
dataloader/__init__.py (0, 2021-01-06)
dataloader/data_augment.py (5494, 2021-01-06)
dataloader/load_data.py (2243, 2021-01-06)
dataloader/loader.py (8950, 2021-01-06)
dataloader/loader_val.py (10363, 2021-01-06)
eval_city/ (0, 2021-01-06)
eval_city/__init__.py (0, 2021-01-06)
eval_city/cocoapi/ (0, 2021-01-06)
eval_city/cocoapi/LuaAPI/ (0, 2021-01-06)
eval_city/cocoapi/LuaAPI/CocoApi.lua (10726, 2021-01-06)
eval_city/cocoapi/LuaAPI/MaskApi.lua (10156, 2021-01-06)
eval_city/cocoapi/LuaAPI/cocoDemo.lua (791, 2021-01-06)
eval_city/cocoapi/LuaAPI/env.lua (436, 2021-01-06)
eval_city/cocoapi/LuaAPI/init.lua (498, 2021-01-06)
eval_city/cocoapi/LuaAPI/rocks/ (0, 2021-01-06)
eval_city/cocoapi/LuaAPI/rocks/coco-scm-1.rockspec (821, 2021-01-06)
eval_city/cocoapi/MatlabAPI/ (0, 2021-01-06)
eval_city/cocoapi/MatlabAPI/CocoApi.m (13117, 2021-01-06)
eval_city/cocoapi/MatlabAPI/CocoEval.m (22360, 2021-01-06)
eval_city/cocoapi/MatlabAPI/CocoUtils.m (16584, 2021-01-06)
eval_city/cocoapi/MatlabAPI/MaskApi.m (4979, 2021-01-06)
eval_city/cocoapi/MatlabAPI/cocoDemo.m (1204, 2021-01-06)
eval_city/cocoapi/MatlabAPI/evalDemo.m (1808, 2021-01-06)
eval_city/cocoapi/MatlabAPI/gason.m (2395, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/ (0, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/gasonMex.cpp (9227, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/gasonMex.mexa64 (38020, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/gasonMex.mexmaci64 (41452, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/getPrmDflt.m (2909, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/maskApiMex.c (5544, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/maskApiMex.mexa64 (21419, 2021-01-06)
eval_city/cocoapi/MatlabAPI/private/maskApiMex.mexmaci64 (23228, 2021-01-06)
eval_city/cocoapi/PythonAPI/ (0, 2021-01-06)
eval_city/cocoapi/PythonAPI/Makefile (199, 2021-01-06)
eval_city/cocoapi/PythonAPI/pycocotools/ (0, 2021-01-06)
... ...

# Center and Scale Prediction (CSP) for pedestrian detection ## Introduction This is the unofficial pytorch implementation of [High-level Semantic Feature Detection: A New Perspective for Pedestrian Detection](https://openaccess.thecvf.com/content_CVPR_2019/papers/Liu_High-Level_Semantic_Feature_Detection_A_New_Perspective_for_Pedestrian_Detection_CVPR_2019_paper.pdf). CSP is an effective and efficient method for pedestrian detector and achieves promising results on the CityPersons dataset. We implement CSP in pytorch based on previous works [offical code (keras)](https://github.com/liuwei16/CSP), [unofficial code (pytorch)](https://github.com/lwpyr/CSP-pedestrian-detection-in-pytorch). Compared with them, our codes have following features: - **Support Apex Mix Precision** - **Support Distributed and Non-distributed training** - **Support more backbones, such as ResNet, DLA-34, HRNet** We obtain much faster training/inference speed (3 hours for 120 epoches using two gpus) and comparable performance. We think CSP is a strong baseline for pedestrian detection, and it still has much room for improvement. We will **continuously update this repo**, and add some useful tricks (e.g. data augmentation) for better performance. ## models | Model | Reasonable | Heavy Occlusion | All | Training time | Link | | ----- | :--------: | :-------------: | :----: | :----------: | :----: | | ResNet-50 | 11.30 | 41.09 | 37.55 | ~5 hours | [BaiduYun(code:v61g)](https://pan.baidu.com/s/1RpuXawJOszBe4BfvIGmZDw) | | DLA-34 | 11.12 | 43.00 | 37.32 | ~3 hours | | | HRNet-18 | 10.24 | 37.72 | 36.15 | ~11 hours | | | HRNet-32| 9.69 | 36.48 | 35.47 | ~13 hours | | |HRNet-32 + [SWA](https://arxiv.org/abs/2012.12645) | 9.66 | **34.61** | 34.86 | | [BaiduYun(code:v61g)](https://pan.baidu.com/s/1RpuXawJOszBe4BfvIGmZDw) | **Note**: Training time is evaluated in two 2080Ti GPUs for 120 epochs. We will further tune some hyperparameters (e.g. learning rate, batchsize) these days, then will release our models. ## Get Start ### Prerequisites - Pytorch 1.2+ - Python 3.6+ - APEX (Install APEX following the [offical instruction](https://github.com/NVIDIA/apex)) ### Installation ````bash git clone git@github.com:ligang-cs/CSP-Pedestrain-detection.git cd CSP-Pedestrian-detection/utils make all ```` ### Data preparation You need to download the [CityPersons](https://github.com/cvgroup-njust/CityPersons) dataset. Your directory tree should be look like this: ````bash $root_path/ ├── images │ ├── train │ └── val ├── annotations │ ├── anno_train.mat │ └── anno_val.mat ```` ### Train and test Please specify the configuration file. #### Distributed training ````bash CUDA_VISIBLE_DEVICES= python -m torch.distributed.launch --nproc_per_node trainval_distributed.py --work-dir ```` #### Non-distributed training ````bash CUDA_VISIBLE_DEVICES= python trainval.py --work-dir ```` #### Test ````bash CUDA_VISIBLE_DEVICES= python test.py --val-path --json-out ```` ## Contact If you have any questions, please do not hesitate to contact Li Gang (gang.li@njust.edu.cn). We also appreciate all contributions to improve this repo. ## Acknowledgement - HRNet codes are folked from [HRNet offical codes](https://github.com/HRNet/HRNet-Semantic-Segmentation/tree/pytorch-v1.1) - DLA-34 codes are folked from [here](https://github.com/ucbdrive/dla) Many thanks to them !

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：