Tool to convert a pre-trained single shot object detection model from MxNet to Caffe.
This tool has been tested with the ssd model trained in MxNet for VOC0712 challenge with the VGG16 front-end.
A dockerfile that creates the environment with Mxnet and Caffe can be found in the docker folder.
- Mxnet version: 0.12.1
- Caffe version: 1.0 with custom layers implemented in Wei Liu's ssd branch of Caffe.
To run the tests launch the run_tests.sh script. Sample results are saved in result.png
