init

README.md

@@ -1,87 +1,58 @@
-# Faster RCNN with PyTorch
-**Note:** I re-implemented faster rcnn in this project when I started learning PyTorch. Then I use PyTorch in all of my projects. I still remember it costed one week for me to figure out how to build cuda code as a pytorch layer :).
-But actually this is not a good implementation and I didn't achieve the same mIOU as the original caffe code. 
+# Faster_R-CNN_pytorch
+- Modifed based on [**longcw/faster_rcnn_pytorch**](https://github.com/daviddao/spatial-transformer-tensorflow) . 
+- Implementation of [**Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks**](https://arxiv.org/abs/1506.01497).  
 
-**This project is no longer maintained. So I suggest:**
-- You can still read and study this code if you want to re-implement faster rcnn by yourself;
-- You can use the better PyTorch implementation by [ruotianluo](https://github.com/ruotianluo/pytorch-faster-rcnn) if you want to train faster rcnn with your own data;
+## Statement
+- Based on **longcw/faster_rcnn_pytorch (python2.7)**, I modified the code to **python3.6** and **numpy 1.13.3**. So the numpy version do not need to change into 1.11.
+- Follow **longcw**'s code, I draw some picture to understand the process of Faster R-CNN.
 
-This is a [PyTorch](https://github.com/pytorch/pytorch)
-implementation of Faster RCNN. 
-This project is mainly based on [py-faster-rcnn](https://github.com/rbgirshick/py-faster-rcnn)
-and [TFFRCNN](https://github.com/CharlesShang/TFFRCNN).
+Thank **longcw**, his code helps me a lot to understand Faster R-CNN. Also the readme files of **longcw** and **ruotianluo** tell me how to build cython models. You can also follow their code and readme file.
 
-For details about R-CNN please refer to the [paper](https://arxiv.org/abs/1506.01497) 
-Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks 
-by Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun.
+**If there is something I can not write in my repertory, please contact me.**
 
-### Progress
+## Environment and Compile
+- python 3.6
+- pytorch 0.3.0
+- numpy 1.13.3
+- opencv 3.3.0
 
-- [x] Forward for detecting
-- [x] RoI Pooling layer with C extensions on CPU (only forward)
-- [x] RoI Pooling layer on GPU (forward and backward)
-- [x] Training on VOC2007
-- [x] TensroBoard support
-- [x] Evaluation
 
-### Installation and demo
-0. Install the requirements (you can use pip or [Anaconda](https://www.continuum.io/downloads)):
+Follow [**longcw/faster_rcnn_pytorch**](https://github.com/longcw/faster_rcnn_pytorch) and [**ruotianluo/pytorch-faster-rcnn**](https://github.com/ruotianluo/pytorch-faster-rcnn).
+
+1. Install requirements (directly copy from [**longcw/faster_rcnn_pytorch/README.md**](https://raw.githubusercontent.com/longcw/faster_rcnn_pytorch/master/README.md)).
 
  ```
  conda install pip pyyaml sympy h5py cython numpy scipy
  conda install -c menpo opencv3
  pip install easydict
  ```
+
+2. Choose arch of GPU type. (directly copy from[**ruotianluo/pytorch-faster-rcnn/README.md**](https://raw.githubusercontent.com/ruotianluo/pytorch-faster-rcnn/master/README.md))
+
+ | 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 |
+
+ modify the parameter **arch** in **faster_rcnn_pytorch/faster_rcnn/make.sh**
+
+3. Build Cython models (directly copy from [**longcw/faster_rcnn_pytorch/README.md**](https://raw.githubusercontent.com/longcw/faster_rcnn_pytorch/master/README.md)).
 
-
-1. Clone the Faster R-CNN repository
- ```bash
- git clone [email protected]:longcw/faster_rcnn_pytorch.git
- ```
-
-2. Build the Cython modules for nms and the roi_pooling layer
  ```bash
  cd faster_rcnn_pytorch/faster_rcnn
  ./make.sh
  ```
-3. Download the trained model [VGGnet_fast_rcnn_iter_70000.h5](https://drive.google.com/open?id=0B4pXCfnYmG1WOXdpYVFybWxiZFE) 
-and set the model path in `demo.py`
-3. Run demo `python demo.py`
-
-### Training on Pascal VOC 2007
-
-Follow [this project (TFFRCNN)](https://github.com/CharlesShang/TFFRCNN)
-to download and prepare the training, validation, test data 
-and the VGG16 model pre-trained on ImageNet. 
-
-Since the program loading the data in `faster_rcnn_pytorch/data` by default,
-you can set the data path as following.
-```bash
-cd faster_rcnn_pytorch
-mkdir data
-cd data
-ln -s $VOCdevkit VOCdevkit2007
-```
-
-Then you can set some hyper-parameters in `train.py` and training parameters in the `.yml` file.
-
-Now I got a 0.661 mAP on VOC07 while the origin paper got a 0.699 mAP.
-You may need to tune the loss function defined in `faster_rcnn/faster_rcnn.py` by yourself.
 
-### Training with TensorBoard
-With the aid of [Crayon](https://github.com/torrvision/crayon),
-we can access the visualisation power of TensorBoard for any 
-deep learning framework.
+## Run
+You can follow [**longcw/faster_rcnn_pytorch**](https://github.com/daviddao/spatial-transformer-tensorflow) .
 
-To use the TensorBoard, install Crayon (https://github.com/torrvision/crayon)
-and set `use_tensorboard = True` in `faster_rcnn/train.py`.
+- **demo.py** You need to download **trained_weights(VGGnet_fast_rcnn_iter_70000.h5)** and load it's weight.
+- **train.py** You need to download **VOC dataset** and ** VGG_imagenet_pretrained_weight(VGG_imagenet.npy)**.
+- **test.py** YOu need to download **VOC dataset**.
 
-### Evaluation
-Set the path of the trained model in `test.py`.
-```bash
-cd faster_rcnn_pytorch
-mkdir output
-python test.py
-```
+## Process of Code
+![](pics/process.png)
 
-License: MIT license (MIT)

data/VOCdevkit2012/readme.md

@@ -0,0 +1,2 @@
+Download VOC dataset.
+The structure of data_dir is data/VOCdevkit2012/VOC2012/[Annotations|ImageSets|...]

data/pretrained_model/readme.md

@@ -0,0 +1 @@
+Download VGG_imagenet.npy

demo.py

@@ -12,7 +12,8 @@ def test():
  # im_file = '/media/longc/Data/data/2DMOT2015/test/ETH-Crossing/img1/000100.jpg'
  image = cv2.imread(im_file)
 
- model_file = '/media/longc/Data/models/VGGnet_fast_rcnn_iter_70000.h5'
+ model_file = 'models/saved_model3/faster_rcnn_100000.h5'
+ # model_file = 'models/VGGnet_fast_rcnn_iter_70000.h5'
  # model_file = '/media/longc/Data/models/faster_rcnn_pytorch3/faster_rcnn_100000.h5'
  # model_file = '/media/longc/Data/models/faster_rcnn_pytorch2/faster_rcnn_2000.h5'
  detector = FasterRCNN()
@@ -27,6 +28,10 @@ def test():
  t = Timer()
  t.tic()
  # image = np.zeros(shape=[600, 800, 3], dtype=np.uint8) + 255
+
+ # dets_np (n_box, 4)
+ # socres_np (n_box, ) [0.9, 0.96, ...]
+ # classes_np (n_box, ) ['dog', 'human', 'people', ...]
  dets, scores, classes = detector.detect(image, 0.7)
  runtime = t.toc()
  print('total spend: {}s'.format(runtime))
@@ -38,8 +43,8 @@ def test():
  cv2.putText(im2show, '%s: %.3f' % (classes[i], scores[i]), (det[0], det[1] + 15), cv2.FONT_HERSHEY_PLAIN,
  1.0, (0, 0, 255), thickness=1)
  cv2.imwrite(os.path.join('demo', 'out.jpg'), im2show)
- cv2.imshow('demo', im2show)
- cv2.waitKey(0)
+ # cv2.imshow('demo', im2show)
+ # cv2.waitKey(0)
 
 
 if __name__ == '__main__':