This repository has collected some useful scripts for object detection, including model transfomer for various framework (e.g darknet2caffe), metric calculator (e.g map).
This module supports the transformation of the detection models trained within different framework. Currently, we have implemented the following features:
- yolov1-caffe
- yolov2-caffe
- yolov3-caffe
Now our tool can be used for the transformation from arbitrary Yolo version to Caffemodel. Note that you can directly use the yolov1(v2)-caffe with the original Caffe version. If you want to use yolov3-caffe, you must add the upsample-layer into the original Caffe and recompile it. You can find the source code of the upsample-layer in the extra_caffe_layers dir. You can get more information in add new layers.