• 1. Adding New Dataset
  • Customize datasets by reorganizing data
  • Customize datasets by mixing dataset
• 2. Custom Data Pipelines
  • Design of Data pipelines
  • Extend and use custom pipelines
• 3. Adding New Modules
  • Customize optimizer
  • Customize optimizer constructor
  • Develop new components
• 4. Training Tricks
  • Different Learning Rate(LR) for Backbone and Heads
  • Online Hard Example Mining (OHEM)
  • Class Balanced Loss