Person detection is a trivial task which I know my detector does fine at anyway. Hence, it doesn't make sense to go through the expensive process of CNN evaluation and GM inference on entire test images. I should make the testing pipeline crop the image around the GT person location and then rescale to GT scale.
Person detection is a trivial task which I know my detector does fine at anyway. Hence, it doesn't make sense to go through the expensive process of CNN evaluation and GM inference on entire test images. I should make the testing pipeline crop the image around the GT person location and then rescale to GT scale.