文件列表:

LICENSE (1067, 2018-11-23)
__pycache__ (0, 2018-11-23)
__pycache__\train.cpython-36.pyc (1550, 2018-11-23)
__pycache__\validate.cpython-36.pyc (1204, 2018-11-23)
dataset (0, 2018-11-23)
dataset\bird (0, 2018-11-23)
dataset\bird\0.jpg (200542, 2018-11-23)
dataset\bird\1.jpg (84435, 2018-11-23)
dataset\bird\2.jpg (69641, 2018-11-23)
dataset\bird\3.jpg (113696, 2018-11-23)
dataset\bird\4.jpg (101244, 2018-11-23)
dataset\bird\5.jpg (63812, 2018-11-23)
dataset\bird\6.jpg (136013, 2018-11-23)
dataset\bird\7.jpg (100340, 2018-11-23)
dataset\bird\8.jpg (126105, 2018-11-23)
dataset\bird\9.jpg (107171, 2018-11-23)
drawrect.py (5880, 2018-11-23)
main.py (7877, 2018-11-23)
model (0, 2018-11-23)
model\DFL.py (2141, 2018-11-23)
model\__pycache__ (0, 2018-11-23)
model\__pycache__\Demo.cpython-36.pyc (1633, 2018-11-23)
model\__pycache__\Googlenet.cpython-36.pyc (3585, 2018-11-23)
model\__pycache__\Googlenet_mobile.cpython-36.pyc (3929, 2018-11-23)
model\__pycache__\inceptionv4.cpython-36.pyc (9160, 2018-11-23)
run.sh (152, 2018-11-23)
screenshot (0, 2018-11-23)
screenshot\introduction1.jpg (44666, 2018-11-23)
screenshot\introduction2.png (25802, 2018-11-23)
screenshot\test.jpg (304297, 2018-11-23)
screenshot\vis_1.jpg (176863, 2018-11-23)
screenshot\vis_2.jpg (169913, 2018-11-23)
screenshot\vis_3.jpg (132863, 2018-11-23)
screenshot\vis_4.jpg (173028, 2018-11-23)
train.py (2537, 2018-11-23)
utils (0, 2018-11-23)
utils\MyImageFolderWithPaths.py (656, 2018-11-23)
utils\__pycache__ (0, 2018-11-23)
... ...

# DFL-CNN : a fine-grained classifier This is a simple pytorch re-implementation of CVPR 2018 [Learning a Discriminative Filter Bank Within a CNN for Fine-Grained Recognition](https://arxiv.org/pdf/1611.09932.pdf). ### Introduction: This work still need to be updated. The features are summarized blow: + Use VGG16 as base Network. + Dataset [CUB-200-2011](http://www.vision.caltech.edu/visipedia/CUB-200-2011.html), you can split **trainset/testset** by yourself.**Or** you can download dataset which has been split directly from [BaiduYun Link](https://pan.baidu.com/s/1JQxa3DYDrM329skC73kbzQ). + This work has been trained on 4 Titan V after epoch 120 with batchsize 56, Now I got best result **Top1 85.140% Top5 96.237%** which is lower than author's. You can download weights from [weights](https://pan.baidu.com/s/1nxI3mV2cOOoMCLpCqg_cjA). + Part FCs is replaced by Global Average Pooling to reduce parameters. + Every some epoches, ten best patches is visualized in **vis_result** directory, you can put images you want to visualize in **vis_img** named number.jpg. + Update: ResNet-101 DFL-CNN and Multi-scale DFL-CNN need to be done. ### Algorithms Introduction:   ### Results and Visualization of ten boxes for discriminative patches: + This work has been trained on 4 Titan V after epoch 120 with batchsize 56, Now I got best result **Top1 85.140% Top5 96.237%** which is lower than author's. You can download weights from [weights](https://pan.baidu.com/s/1nxI3mV2cOOoMCLpCqg_cjA). If use TenCrop transform in code, result can improve further. + Test Results: + Visualization: ### Usage: + Download dataset, you can split trainset/valset by yourself ``` wget http://www.vision.caltech.edu/visipedia-data/CUB-200-2011/CUB_200_2011.tgz ``` + Or you can directly get it from [BaiduYun Link](https://pan.baidu.com/s/1JQxa3DYDrM329skC73kbzQ) + Then link original dataset to our code root/dataset ``` ln -s ./train path/to/code/dataset/train ln -s ./test path/to/code/dataset/test ``` + Finally, Train and Test. + Check you GPU resources and modify your run.sh. ``` sh run.sh ``` ### Note: 1. Visualization of ten best boxes is saved in **vis_result/**, img you want to visualize should be put in **vis_img/**. 2. Weight(checkpoint.pth.tar, model_best.pth.tar) is in **weight/**. 3. Loss info is saved in **log/**.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：