Change class number

[Nietram]

Hi, I don't understand really well where you setup the squares and nb of classes. Is there some needs to change that into the utils.py function?

[Altimis]

You dont have to setup the number of classes. This code is the continuation of the code example presented in the main Mask-RCNN repo. You just compute the predictions for each images, assign each prediction (classe) to its actual ground-truth using two ordered vectors (gt_tot and pred_tot), and classically compute the confusion matrix for the entire dataset. These two vectors will contain in each element the ground-truth classe for each object in the dataset (the vector gt_tot) and the corresponding predicted classe (the vector pred_tot).