I have labeled a set of images with both pixel level label (for semantic segmentation) and bounding boxes (for object detection). I loaded the data directory and the data (gTruth.mat), but I cannot get the dataset (images with bounding box info) to load as trainingData. I have issues defining the bounding box datastore also. Is this because I need to only label the images with bounding boxes and save the ground truth separately from the pixel label set?
I was hoping to use the set for mask r cnn so I did the pixel and bounding box labels together. Semantic segmentation alone didn't work for all the classes. Now I want to trying object detection before moving to incident segmentation.