I am working on a project that uses object detection. I have logo images that need to be detected in a video. I am doing this in keras. I followed this blog to convert the yolo weights to a keras model.
Now I would like to train this keras model with my dataset (transfer learning). I got right dataset with xmin, ymin, xmax, and ymax.
My Approach: remove the last layer and use the sigmoid function to define the classes. However, the above link says that this keras model predicts both boundary boxes and classes. I am not sure how to achieve these boundary boxes from the model itself.