DIP-proj-DepthEstimation

Directories

• ./model

  • block.py
    • define the encoder(now only resnet50), mff, decoder and refinement block, which are modules of the model
  • model.py
    • define the model, utilizing the blocks in block.py
  • resnet_module.py
    • self-defined resnet for encoding blocks
      • credit to https://github.com/JunjH/

• nyu2_train

  • only part of the dataset, used to check whether the code is runnable

• load_data.py

  • load the data by making pairs of path, the output is a list of pair [x_tr_path, y_tr_path]

• loss.py

  • define the gradient computation and loss computation

• utils.py

  • currently, only loading and saving the params of model is defined

• train.py

  • define the training procedure