This project contains analysis of various techniques such as separable convolutions(MobileNet) and grouped convolutions (ShuffleNet) using concatenation and addition for skip connections
Instructions :
yolodws_concat.py contains the code for depthwise separable convolution with concatenation
yolodws_concat_predict.py contains the code for prediction
extras: yolodws_add.py contains the code for depthwise separable convolution with addition basicyolo.py contains the code of actual un optimized network shuffle.py contains the code for grouped convolutions using shuffle technique.
References :
https://github.com/experiencor/keras-yolo2/blob/master/Yolo%20Step-by-Step.ipynb