agv-semseg-pipeline's Introduction

AGV DL Segmentation Pipeline

Structure

├── config  
│   ├── vanilla_enet.json
├── data
├── dataloader
│   ├── __init__.py
│   ├── cityscapes.py
├── models
│   ├── __init__.py
│   ├── enet.py
├── utils
│   ├── __init__.py
│   ├── augmentation.py
│   ├── config.py
│   ├── device.py
│   ├── losses.py
│   ├── metric.py
│   ├── preprocessing.py
│   ├── saving.py
│   ├── scripts.py
│   ├── tester.py
│   ├── trainer.py
│   ├── wandb_utils.py
├── __init__.py
├── README.md
├── run.py
├── test.py
├── train.py

Usage

Downloading Data

wget --keep-session-cookies --save-cookies=cookies.txt --post-data 'username=username&password=pass&submit=Login' https://www.cityscapes-dataset.com login/
wget --load-cookies cookies.txt --content-disposition https://www.cityscapes-dataset.com/file-handling/?packageID=1
unzip -q -o gtFine_trainvaltest.zip
wget --load-cookies cookies.txt --content-disposition https://www.cityscapes-dataset.com/file-handling/?packageID=3
unzip -q -o leftImg8bit_trainvaltest.zip

Training and Testing

python3 run.py [-h] [--config CONFIG_FILE] [--mode MODE]
                    [--wandb_id "WANDB_API_KEY"]

Results

Mean Accuracy and IoU

Experiment	Training Mean IoU	Validation Mean IoU	Training Mean Accuracy	Validation Mean Accuracy
Weighted CutMix ENet	0.62	0.57	0.66	0.63
CutMix ENet	0.60	0.52	0.73	0.69
Vanilla ENet	0.56	0.48	0.70	0.66

Segmented Images

Contributors

This repository is maintained by AGV (IIT Kharagpur)

Refrences

The Cityscapes Dataset for Semantic Urban Scene Understanding: Link
ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation: Link
CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features: Link

Recommend Projects