Closed LevinJ closed 11 months ago
@LevinJ I'm sorry that it only supports batch_size=1 for now and you can split it for 4 or more images evaluation in a iteration.
@DRosemei , Thanks for your kind reply. Yes, I later also realized we could use the eval script to do the evaluation. Truly appreciate your generous act of opening sourcing this nice project, which is very helpful and inspiring for me. Great work!
After enabling the eval feature (via eval: True) in local_nusc.yaml, evaluation relevant scripts will be executed at the end of the training, and below numpy size match error would occur at the
blend_image = cv2.addWeighted(gt_image, 0.5, vis_seg, 0.5, 0)
line.gt_image is of shape (4, 225,800,3), while vis_seg is of (225, 800,3)