Refactor
6/5: changed the myKit.py,but not updata the requirements.txt
6/7 updata the myKit.py,requirements.txt,mymodel.py add Animator.py,accuracy.ipynb for evaluating the output of model add dir "animator" for saving the loss map
6/9 Training MAE loss is 7.37, when we delete the MMCA module , resualt is seem to be worth.So we spurpose to add the MMCA module and freeze the resnet feature net
6/13 Freeze the MMCA module
6/17 Due to the detached-training is not good, we try to use the original RseNet50 to training.Subsequencely，we don't sample the training-dataset.