This repository is a starter-code setup for Resource Constraint Recyclable Waste Segmentation project. It provides the code base for training ENet on the ReSort dataset for binary class segmentation.
dataset.zip
into the project folder.__C.DATA.DATA_PATH
in config.py
.python train.py
command to train the model.train.py
also provides the flexibility of either training the entire model (encoder + decoder) or just the encoder which can be performed by changing __C.TRAIN.STAGE
in config.py
.model.py
contains the model definition of ENet. To train on newer models such as the ICNet model definition of such models needs to be added inmodel.py
.validate
function of train.py
and dataloader class resortit
needs to be modified accordingly.FLOPS
and # of trainable model parameters
.