pursu / YOLO3-TF

YOLO v3 re-implementation

YOLO3-TF

YOLO v3 re-implementation, and our code is based stronger-yolo, a huge thank to him.

Use backbone with VGG-16 and Mobilenet V2

Usage

1. clone this repository

   git clone https://github.com/fanbinqi/YOLO3-TF.git

2. prepare data
   (1) download datasets
   Create a new folder named data and then create a new folder named VOC in the data/.
   Download VOC 2012_trainval 、VOC 2007_trainval 、VOC 2007_test, and put datasets into data/VOC, name as 2012_trainval、2007_trainval、2007_test separately.
   The file structure is as follows:
   |--YOLOV3-TF
   |--data
   |--|--VOC
   |--|--|--2012_trainval
   |--|--|--2007_trainval
   |--|--|--2007_test
   (2) convert data format
   You should set DATASET_PATH in config.py to the path of the VOC dataset and then
   

   python voc_annotation.py

3. prepare initial weights
   (1) yolov3 Download YOLOv3-608.weights firstly, put the yolov3.weights into yolov3_to_tf/, and then

   cd yolov3_to_tf
   python3 convert_weights.py --weights_file=yolov3.weights --dara_format=NHWC -- ckpt_file=./saved_model/yolov3_608_coco_pretrained.ckpt
   cd ..
   python rename.py

   (2) mobilenet v2 if want to train model with MoBileNet V2, the re-trained model is in our repo ./weights

   put the initial weight into `weights/`.

4. Train
   

   python train.py

5. Test
   

   python test.py
   cd mAP
   python main.py

If you are interested in this project, please QQ me (374873360)