# DetNet_pytorch **Repository Path**: usefordev/DetNet_pytorch ## Basic Information - **Project Name**: DetNet_pytorch - **Description**: An implementation of DetNet: A Backbone network for Object Detection. - **Primary Language**: Unknown - **License**: MIT - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2020-08-29 - **Last Updated**: 2020-12-19 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README An implementation of [DetNet: A Backbone network for Object Detection](https://arxiv.org/abs/1804.06215). Due to the short time, I only trained and tested on pascal voc dataset. It proved that the performance of detnet59 is indeed better than FPN101. ## Introduction Firstly, I spent about one week training detnet59 on the ImageNet dataset .The classification performance of detnet59 is a little better than the original resnet50. Then i used the pretrained detnet59 to train and test on pascal voc. Based on [**FPN_Pytorch**](https://github.com/guoruoqian/FPN_Pytorch/), i change FPN101 to detnet59. **Update** **Adding soft_nms.** **Without requiring any re-training of existing models.** You only need to use soft_nms during testing to bring performance improvements. ## Benchmarking I benchmark this code thoroughly on pascal voc2007 and 07+12. Below are the results: 0). ImageNet(test on validation dataset) | backbone | Top1 error | | ------------------------------ | ---------- | | pytorch resnet50 | 23.9 | | detnet59 in this code | 23.8 | | detnet59 in the original paper | 23.5 | 1). PASCAL VOC 2007 (Train/Test: 07trainval/07test, scale=600, ROI Align) | model（FPN） | GPUs | Batch Size | lr | lr_decay | max_epoch | Speed/epoch | Memory/GPU | mAP | | ------------------------------------------------------------ | --------------- | ---------- | ---- | -------- | --------- | ----------- | ---------- | ---- | | ResNet-101 | 1 GTX 1080 (Ti) | 2 | 1e-3 | 10 | 12 | 1.44hr | 6137MB | 75.7 | | [DetNet59](https://www.dropbox.com/home/DetNet/PASCAL%20VOC%202007?preview=fpn_1_8_5010.pth) | 1 GTX 1080 (Ti) | 2 | 1e-3 | 10 | 12 | 1.07hr | 5412MB | 75.9 | 2). PASCAL VOC 07+12 (Train/Test: 07+12trainval/07test, scale=600, ROI Align) | model（FPN） | GPUs | Batch Size | lr | lr_decay | max_epoch | Speed/epoch | Memory/GPU | mAP | | ------------------------------------------------------------ | --------------- | ---------- | ---- | -------- | --------- | ----------- | ---------- | ---- | | ResNet-101 | 1 GTX 1080 (Ti) | 1 | 1e-3 | 10 | 12 | 3.96hr | 9011MB | 80.5 | | [DetNet59](https://www.dropbox.com/home/DetNet/PASCAL%20VOC%2007%2B12?preview=fpn_1_7_33101.pth) | 1 GTX 1080 (Ti) | 1 | 1e-3 | 10 | 12 | 2.33hr | 8015MB | 80.7 | | ResNet-101(**using soft_nms when testing**) | 1 GTX 1080 (Ti) | \ | \ | \ | \ | \ | \ | 81.2 | | DetNet59(**using soft_nms when testing**) | 1 GTX 1080 (Ti) | \ | \ | \ | \ | \ | \ | 81.6 | ## Preparation First of all, clone the code ``` git clone https://github.com/guoruoqian/DetNet_Pytorch.git ``` Then, create a folder: ```shell cd DetNet_Pytorch && mkdir data ``` ### prerequisites - Python 2.7 or 3.6 - Pytorch 0.2.0 or higher（not support pytorch version >=0.4.0） - CUDA 8.0 or higher - tensorboardX ### Data Preparation - VOC2007: Please follow the instructions in [py-faster-rcnn](https://github.com/rbgirshick/py-faster-rcnn#beyond-the-demo-installation-for-training-and-testing-models) to prepare VOC datasets. Actually, you can refer to any others. After downloading the data, creat softlinks in the folder data/. - VOC 07 + 12: Please follow the instructions in [YuwenXiong/py-R-FCN](https://github.com/YuwenXiong/py-R-FCN/blob/master/README.md#preparation-for-training--testing) . **I think this instruction is more helpful to prepare VOC datasets.** ### Pretrained Model  You can download the detnet59 model which i trained on ImageNet from: - detnet59: [dropbox](https://www.dropbox.com/home/DetNet?preview=detnet59.pth)，[baiduyun](https://pan.baidu.com/s/14_ztsAKcrZGb4nnm8aCMyQ) Download it and put it into the data/pretrained_model/. ### Compilation As pointed out by [ruotianluo/pytorch-faster-rcnn](https://github.com/ruotianluo/pytorch-faster-rcnn), choose the right `-arch` in `make.sh` file, to compile the cuda code: | GPU model | Architecture | | :------------------------- | :----------: | | TitanX (Maxwell/Pascal) | sm_52 | | GTX 960M | sm_50 | | GTX 1080 (Ti) | sm_61 | | Grid K520 (AWS g2.2xlarge) | sm_30 | | Tesla K80 (AWS p2.xlarge) | sm_37 | Install all the python dependencies using pip: ```shell pip install -r requirements.txt ``` Compile the cuda dependencies using following simple commands: ```shell cd lib sh make.sh ``` It will compile all the modules you need, including NMS, ROI_Pooing, ROI_Align and ROI_Crop. The default version is compiled with Python 2.7, please compile by yourself if you are using a different python version. ## Usage train voc2007: ```shell CUDA_VISIBLE_DEVICES=3 python3 trainval_net.py exp_name --dataset pascal_voc --net detnet59 --bs 2 --nw 4 --lr 1e-3 --epochs 12 --save_dir weights --cuda --use_tfboard True ``` test voc2007: ```shell CUDA_VISIBLE_DEVICES=3 python3 test_net.py exp_name --dataset pascal_voc --net detnet59 --checksession 1 --checkepoch 7 --checkpoint 5010 --cuda --load_dir weights ``` **using soft_nms when testing**: ```shell CUDA_VISIBLE_DEVICES=3 python3 test_net.py exp_name --dataset pascal_voc --net detnet59 --checksession 1 --checkepoch 7 --checkpoint 5010 --cuda --load_dir weights --soft_nms ``` Before training voc07+12, you can must set ASPECT_CROPPING in detnet59.yml False, or you will encounter some error during the training. train voc07+12: ```shell CUDA_VISIBLE_DEVICES=3 python3 trainval_net.py exp_name2 --dataset pascal_voc_0712 --net detnet59 --bs 1 --nw 4 --lr 1e-3 --epochs 12 --save_dir weights --cuda --use_tfboard True ``` ### TODO - Train and test on COCO