Detect the Robosapien in an image using darkflow.
Install darkflow from here. Ensure to go through the readme thoroughly. You can build for both python2 and python3.
The ./cfg and ./bin directories in this repository contain the cfg and weight file. You can use these files to train a single class object detector from scratch.
./built_graph contains the built graph as a protobuf (.pb) file and the meta file (.meta). See the darkflow repository for more on these files.
Integrate these folders with their respective darkflow directories. You will have to create the ./built_graph directory in darkflow, just copy and paste it.
Test the neural network with your own image containing the robosapien. In your darkflow repository run;
flow --pbLoad built_graph/tiny-yolo-voc-1c.pb --metaLoad built_graph/tiny-yolo-voc-1c.meta --imgdir sample_img/
Check darkflow/sample_img/out to see the results.
You can train on your own robosapien dataset using the built graph. From darkflow/ run;
flow --train --model cfg/tiny-yolo-voc-1c.cfg --load 3625
Ensure to go through the darkflow repository for more info.