search

asanakoy / deeppose_tf

DeepPose implementation on TensorFlow. Original Paper http://arxiv.org/abs/1312.4659
Other
144 stars 61 forks source link

Train error does not decrease. #6

Closed jjzhang10 closed 7 years ago

jjzhang10 commented 7 years ago

Hi, I use the train_lsp_alexnet_imagenet.sh to train. But it seems the training error get stucks to [0.1 0.11] and cannot decrease any more.