https://github.com/jahongir7174/deeplab-pt

DeepLabV3 plus implementation using PyTorch
https://github.com/jahongir7174/deeplab-pt

deeplab-v3-plus efficientnetv2 pytorch training

Last synced: about 2 months ago
JSON representation

DeepLabV3 plus implementation using PyTorch

• Host: GitHub
• URL: https://github.com/jahongir7174/deeplab-pt
• Owner: jahongir7174
• License: apache-2.0
• Created: 2021-04-13T07:05:59.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2021-07-29T05:10:00.000Z (almost 5 years ago)
• Last Synced: 2025-12-30T08:57:29.750Z (5 months ago)
• Topics: deeplab-v3-plus, efficientnetv2, pytorch, training
• Language: Python
• Homepage:
• Size: 11.7 KB
• Stars: 4
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# DeepLabV3+

[DeepLabV3+](https://arxiv.org/pdf/1802.02611.pdf) implementation for Semantic Segmentation using PyTorch

#### Train

* Run `python main.py` for training

#### Dataset structure (similar to CamVid dataset)

    ├── Dataset folder 

        ├── train

            ├── 1111.png

            ├── 2222.png

        ├── train_labels

            ├── 1111_L.png

            ├── 2222_L.png

        ├── class_dict.csv

 

#### Note 

* default feature extractor is [EfficientNetV2-S](https://arxiv.org/pdf/2104.00298.pdf)

* changing configuration of training, change parameters in `utils/config.py`

* default loss function is `weighted cross entropy`