Hi, I need to use ImagePosition, but I don't find a way to give images to learn what have to recognize.
There are only examples with "concept", "word" or Entity.
I tried with Classify to use a ClassifierFunction but no success, it seems not a valid object category.
Thanks a lot for your time.
Have a good day.
What are you trying to recognize in the images? Please provide some sample images. Do you have an accurately labeled dataset of images?
Thanks Rohit Namjoshi for your time.
I am trying to get skin moles positions from skin frames. I am giving samples taken directly from the same frames of the video, so the recognition should be faster and easier.
My next step could be subdivide frames and run every subframes on Classify and get positions, but I think it could be much heavier than use neural network (I still don’t know how works) or trying find some built-in concept similar to skin moles. If you have some suggestion.
ImagePosition is using a pre-trained models, so you don't have to provide examples to train it. If you want to recognize an object that ImagePosition cannot recognize, then you will have to build your own model, but first provide more details on what you want to recognize, maybe it is already defined somewhere.
Thanks Ahmed Elbanna, it was first answer I was looking for.
One approach would be to use sub-images, training a classifier to recognize presence vs. absence of a mole ( or part thereof). Once trained e.g. With 'Classify', actual testing would flag sub-images containing moles, and the location of the sub-image within the full image would indicate body position of that mole.
Thanks a lot Daniel Lichtblau, I think it will be my first solution