lyfsgithud/yolo3-training-keras-master

YOLOv3 training with own data on GPU. A Keras implementation of YOLOv3

yolo3-training-keras-master

---

Introduction

1. use YOLOv3 training with own data
2. Compile YOLOv3 training on GPU
3. A Keras implementation of YOLOv3 detection inspired by qqwweee/keras-yolo3.

---

YOLOv3 training

see YOLOv3训练自己的数据 for details.

YOLOv3 training with GPU

see darknet YOLO 编译使用GPU for details.

Keras implementation of YOLOv3

training with own data(rock climbers,classes of male and female)

1. Download trained YOLOv3 weights from baidu yun.
2. Convert the Darknet YOLO model to a Keras model.
3. Run YOLO detection.

python convert.py ./cfg/yolov3.cfg yolov3.weights yolo.h5
python yolo.py

PS:

1. change value of 'self.score' in yolo.py to adjust detection threshold
2. In my case, I trained 400 epoch with 200 pictures,it is far from enough to obtain a good test result.
3. some examples: