YOLOV2 - Caffe Converter

This is an implementation of YOLO v2 model to Caffemodel converter, thanks to Duangenquan

Supported models

Tiny YOLO v2

Anchor box (Region layer) is implemented in the parser / applications. This converter does not support the Route and Reorg layer (yet).

..* Note: yolov2Tiny20 in the yolomodels is actually tiny-yolo-voc from YOLO original web

How to use it:

1. Prepare the pre-trained YOLO models (config file .cfg and pre-trained weights .weights).

2. Convert the config file using create_yolo_prototxt.py python create_yolo_prototxt.py <cfg_file> <prototxt_output>

3. Convert the pre-trained weights using create_yolo_caffemodel.py python create_yolo_caffemodel.py -m <model_file> -w <yoloweight_file> -o <caffemodel_output>

Warning!

Please note that unlike Darknet, Caffe does not support default padding in their pooling layers. Depends on your model, you might find a difference in the output size. In our case of tiny-yolo-voc, Darknet produces 13x13x125 output whilst Caffe produces 12x12x125.