DeepLab V3+ is a state-of-the-art model for semantic segmentation.
This repository contains a PyTorch implementation of DeepLab V3+ trained for full driving scene segmentation tasks.
- Base DeepLab model
- DeepLab decoder module
- Xception feature extractor backbone
- Dataloaders, train script, metrics
- Data augmentation pipeline
- Upload pretrained weights and results
- Planned extensions
- Mixed precision training
- Faster, MobileNet-v2 backbone
- Pretraining all backbones on ImageNet
- Dataloader for City-scapes
- Dataloader for Baidu apollo dataset
- Setup inference mode