search

qiexing / face-landmark-localization

cnn network predict face landmarks (68 points) and head pose (3d pose, yaw,roll,pitch).
410 stars 184 forks source link

Multiple deploy.prototxt files #22

Open sanjanajainsj opened 7 years ago

sanjanajainsj commented 7 years ago

Hello, Thank you for the model and codes. I am a bit confused. There are two deploy files provided. The model/deploy.prototxt consists of a model which gets divided at the fully connected layer fc7 to two parts: 68-point and poselayer. However, the train/deploy.prototxt consists of a model which gets divided after the relu4 layer to be passed onto conv5 and conv5_b, and so is the train/train_val.prototxt. What are the differences in results for these and which one do you recommend? Also, we just have one train_val.prototxt. Is the caffemodel trained on a file similar to model/deploy.prototxt? Thank you in advance.

qiexing commented 7 years ago

Yes, you are right. These are two different network structures. The difference is how many layers these two tasks are shared. The latter network structure will cost more feed forward time.

zhouxiaoxu commented 7 years ago

qiexing 你好，你提供的model是使用 train/train_val.prototxt训练的吗？如果是，那使用 train/train_val.prototx训练的模型可以在使用model/deploy.prototxt做预测吗？