Converting generated disparity npy to png format

[Mamoanwar97]

I used the generate_disp script found in pseudo_lidar repo, I used it in converting the point cloud of Kitti object detection 7481 examples in order to have large data to train the model on, but in order to train i need to convert the images from npy format to png, i tried many ways

this one of them

disp = np.load(args.main_dir + '/' + fn)
disp_map = (disp_map - np.min(disp_map)) / (np.max(disp_map) - np.min(disp_map))
disp_map = (disp_map*255).astype(np.uint8)

it actually output a good looking image but when i start the training, the results are soooo bad, almost nothing is correct so any idea of the right way to convert these npy files to images i can train on?