GitHub Link: https://github.com/Jhuangsp/VRDL_HW4

• train.py: Main file used to train the model
• eval.py: Main file used to evaluate the test data
• yolact.py: YOLACT model definition
• backbone.py: Backbone model definition used by YOLACT model
• TA_utils.py: Utils provide by TA, used to convert binary mask to rle format.
• data: Storing dataset, configurations (e.g. data path, model select, iterations...).
• layers: Loss functions definition, output utils, box utils.
• utils: Augmentations, logger, timer..., etc.

dbolya/yolact You Only Look At CoefficienTs A simple, fully convolutional model for real-time instance segmentation.