ValueError: Layer #7 (named "bn2a_branch2a"), weight <tf.Variable 'bn2a_branch2a/gamma:0' shape=(256,) dtype=float32_ref> has shape (256,), but the saved weight has shape (64,).

liuwei16 / CSP

High-level Semantic Feature Detection: A New Perspective for Pedestrian Detection, CVPR, 2019

755 stars 191 forks source link

ValueError: Layer #7 (named "bn2a_branch2a"), weight <tf.Variable 'bn2a_branch2a/gamma:0' shape=(256,) dtype=float32_ref> has shape (256,), but the saved weight has shape (64,). #72

Open zhijianglu opened 5 years ago

zhijianglu commented 5 years ago

when i load the pre-trained cityperson model from author ,i met this problem, seems like the model is wrong, have everyone met this

SouleTraveler commented 4 years ago

@zhijianglu Have you sovlee it? I met the similar problem as you

Nexus01 commented 4 years ago

I have a similar problem: ValueError: Dimensions must be equal, but are 256 and 160 for 'P3norm/mul' (op: 'Mul') with input shapes: [?,?,?,256], [160]. Is it possible that the size of pictures don't match the size_train?

quhb2455 commented 4 years ago

@Nexus01 Have you figured out your problem? totally same problem as me.