The paper mentioned that it first trained on WFLW dataset with 98 points landmark and then use the weight as pretrain and finetune on 300w dataset with 68 points landmark. Could you share script how to use pretrain on 98 points that ignore the last fc layer (68 points) thanks
The paper mentioned that it first trained on WFLW dataset with 98 points landmark and then use the weight as pretrain and finetune on 300w dataset with 68 points landmark. Could you share script how to use pretrain on 98 points that ignore the last fc layer (68 points) thanks