possible BUG about the noralization of PoolNet

[WuGuommming]

Thank you for sharing!
I like your work and im trying to understand some details.

I notice that you directly copy and paste the normalization parameters np.array((104.00699, 116.66877, 122.67892)) from PoolNet/dataset to your flow_estimate_saliency function.
However, the dataset of PoolNet is read by cv2.imread(), which the channel order is BGR. It seems doesn't match your RGB data？