So I am tiring to figure this out but keep running into a wall. So what I want to do is create a heatmap from a 2D image based on the key points generated from the MPII dataset. I have found a CSV file with all the keypoints extracted but they are for only one person in the scene. Another issue is what do I used as the data and what do I used as the label. This has been a little confusing for me on where to go.
So my thinking is I have two inputs: 1 is the image and 2 is the keypoints. The labels will be the keypoint name like r_ankel etc...
My issue with this approach is that 1: the images may have more than 1 person in the photo but only one set of keypoint for a particular person. I do not know how to crop the image without throwing off the key points and the size of the person changes and there are not bounding boxes in the CSV that I can go off of.
Any ideas on how to approach this will be appreciated.
Oh fyi I am doing this in Keras.