search

wzjialang / MFPNet

[Remote Sensing] PyTorch implementation for "Remote Sensing Change Detection Based on Multidirectional Adaptive Feature Fusion and Perceptual Similarity"
MIT License
31 stars 6 forks source link

question about training process #5

Open ccw12345 opened 3 years ago

ccw12345 commented 3 years ago

I am confused about your training process. you load the pretrained model，i want to know the pretrained model and other parts of the model are independently trained or freeze the pretrained model，only train the other parts of the model？

wzjialang commented 3 years ago

In our paper, we use SE-ResNet50 as our backbone and initialize it with weights pretrained on ImageNet (available at https://github.com/moskomule/senet.pytorch/releases/download/archive/seresnet50-60a8950a85b2b.pkl). After that, we trained our whole model end-to-end by using the remote sensing change detection dataset.