Pedestrian-Detection-using-YOLO-with-D-IoU-and-C-IoU

This paper proposes a method to improve the performance of pedestrian detection. The method is based on the You Only Look Once (YOLO) algorithm and the improved Intersection over Union (IoU) loss function.

ENVIROMENT

Python 3.5.2 or more
Keras 2.1.5
tensorflow 1.6.0 or more

Structure

model_data------|---voc_annotation.py               # run before training voc dataset
                |---coco_annotation.py              # run before training coco dataset
                |---yolo_weights.h5                 # pre-trained weight file [over 25mb, not uploaded]
                |---yolo_hao.h5                     # trained YOLO model      
                |---xxx_anchor and classess.txt     # traine different dataset  

Test------------|---result_imges                    # store tested images
                |---test_imges                      # images for testing
                |---yolo_test.py                    # test basic model performance for DEMO

yolo3-----------|---model.py                        # modify the training strategy
                |---utils.py                        # bottleneck training added in addtional 
                 
convert.py                                          # Convert the Darknet YOLO model to a Keras model
darknet53.cfg                                       # orignal network from Joseph
kmeans.py                                           # k-means clustering and regression
train.py                                            # train specific model
yolo.py                                             # detecting configuration in images and videos
yolov3.cfg                                          # Converted model from Darknet
yolo_video.py                                       # more command line option parsing

####  Quick Test

1. put test images to /Test/test_imges
2. run yolo_test.py
3. result stored in /Test/result_imges
       |||||
       vvvvv   
==============================RUN IN TERMINAL==============================
cd ./Test
python3 yolo_test.py
===========================================================================


####  Introduction

A Keras implementation of YOLOv3 with modifiled loss function(Tensorflow backend).


####  Quick Start

1. Download YOLOv3 weights from [YOLO website](https://pjreddie.com/darknet/yolo/).
2. Convert the Darknet YOLO model to a Keras model. 
3. Run YOLO detection.
       |||||
       vvvvv
==============================RUN IN TERMINAL==============================
wget https://pjreddie.com/media/files/yolov3.weights
python3 convert.py yolov3.cfg yolov3.weights model_data/yolo.h5
python3 yolo_video.py [OPTIONS...] --image, for image detection mode, OR
python3 yolo_video.py [video_path] [output_path (optional)]
===========================================================================

For another version of YOLO, just do in a similar way, just specify model path and anchor path with `--model model_file` and `--anchors anchor_file`.

Usage

Use --help to see usage of yolo_video.py:

usage: yolo_video.py [-h] [--model MODEL] [--anchors ANCHORS]
                     [--classes CLASSES] [--gpu_num GPU_NUM] [--image]
                     [--input] [--output]

positional arguments:
  --input        Video input path
  --output       Video output path

optional arguments:
  -h, --help         show this help message and exit
  --model MODEL      path to model weight file, default model_data/yolo.h5
  --anchors ANCHORS  path to anchor definitions, default
                     model_data/yolo_anchors.txt
  --classes CLASSES  path to class definitions, default
                     model_data/coco_classes.txt
  --gpu_num GPU_NUM  Number of GPU to use, default 1
  --image            Image detection mode, will ignore all positional arguments

MultiGPU usage: use --gpu_num N to use N GPUs. It is passed to the Keras multi_gpu_model().

Training

Generate your own annotation file and class names file.
One row for one image;
Row format: image_file_path box1 box2 ... boxN;
Box format: x_min,y_min,x_max,y_max,class_id (no space).
For VOC dataset, try python voc_annotation.py
Here is an example:
```
path/to/img1.jpg 50,100,150,200,0 30,50,200,120,3
path/to/img2.jpg 120,300,250,600,2
...
```
Make sure you have run python convert.py -w yolov3.cfg yolov3.weights model_data/yolo_weights.h5
The file model_data/yolo_weights.h5 is used to load pretrained weights.
Modify train.py and start training.
python3 train.py
Use your trained weights or checkpoint weights with command line option --model model_file when using yolo_video.py Remember to modify class path or anchor path, with --classes class_file and --anchors anchor_file.

If you want to use original pretrained weights for YOLOv3:
1. wget https://pjreddie.com/media/files/darknet53.conv.74
2. rename it as darknet53.weights
3. python convert.py -w darknet53.cfg darknet53.weights model_data/darknet53_weights.h5
4. use model_data/darknet53_weights.h5 in train.py

Thanks the author for sharing the Python YOLOv3 source code, the url: https://github.com/qqwweee/keras-yolo3

jianyi324/Pedestrian-Detection-using-YOLO-with-D-IoU-and-C-IoU

Pedestrian-Detection-using-YOLO-with-D-IoU-and-C-IoU

ENVIROMENT

Structure

Usage

Training