---
license: mit
library_name: segmentation_models_pytorch
pipeline_tag: image-segmentation
tags:
- semantic-segmentation
- microstructure
- materials-science
- pytorch
- deeplabv3plus
---

# DeepLabV3+ for Quantitative Microstructure Segmentation

## Model Description

This repository contains seven dataset-specific DeepLabV3+ checkpoints used
in a benchmark of microstructure segmentation methods. The implementation is
from `segmentation_models_pytorch` with a ResNet-50 encoder initialized from
ImageNet weights. Inputs are resized to 512 x 512 pixels and predictions are
restored to the original image size with nearest-neighbor interpolation.

Source code:
https://github.com/WUT-AI-AI4Mat/Segmentation-methods-evaluation-for-quantitative-microstructure-analysis

## Checkpoints

| Dataset | Task | Output channels | File |
| --- | --- | ---: | --- |
| Aachen-Heerlen | Binary segmentation | 1 | `checkpoints/Aachen-Heerlen/DeepLabV3P_Best_Model.pth` |
| EMPS | Binary segmentation | 1 | `checkpoints/EMPS/DeepLabV3P_Best_Model.pth` |
| Grain | Binary segmentation | 1 | `checkpoints/Grain/DeepLabV3P_Best_Model.pth` |
| EBC | 3-class segmentation | 3 | `checkpoints/EBC/DeepLabV3P_Best_Model.pth` |
| Super | 3-class segmentation | 3 | `checkpoints/Super/DeepLabV3P_Best_Model.pth` |
| MetalDAM | 5-class segmentation | 5 | `checkpoints/MetalDAM/DeepLabV3P_Best_Model.pth` |
| UHCS | 7-class segmentation | 7 | `checkpoints/UHCS/DeepLabV3P_Best_Model.pth` |

Binary checkpoints produce one foreground logit; background is represented by
the complementary binary label. Multiclass counts include background.

## Training Details

- Input size: 512 x 512
- Batch size: 32
- Epochs: 500
- Optimizer: AdamW
- Learning rate: 0.0003
- Weight decay: 0.001
- Scheduler: CosineAnnealingLR with minimum learning rate 0.000001
- Early-stopping patience: 50
- Binary objective: Dice and BCEWithLogits
- Multiclass objective: Dice and cross-entropy

## Usage

Example for binary EMPS segmentation:

```bash
hf download NAMESPACE/microstructure-deeplabv3plus checkpoints/EMPS/DeepLabV3P_Best_Model.pth --local-dir weights/deeplabv3plus
python "CNN/DeepLabV3+/test.py" \
  --dataset-root /path/to/EMPS \
  --checkpoint weights/deeplabv3plus/checkpoints/EMPS/DeepLabV3P_Best_Model.pth \
  --output-dir results/deeplabv3plus_emps
```

For EBC, Super, MetalDAM, and UHCS, use
`CNN/DeepLabV3+/test_multi.py` and pass the matching `--num-classes` value.

## Evaluation

The evaluation scripts save original-resolution masks and report mIoU, Dice,
precision, recall, accuracy, HD95, Hausdorff distance, NSD, MAE, MBSS, and
MBSS_add through `Myutils/metrics.py`. Numerical benchmark results will be
linked after the associated paper becomes publicly available.

## Intended Use and Limitations

These checkpoints are intended for research reproduction and comparison on
the named microstructure datasets. They are not general-purpose segmentation
models. Domain shift in imaging, magnification, materials, or annotation
policy may reduce accuracy. They are not validated for safety-critical or
industrial quality-control decisions.

## License and Citation

The released experiment files are provided under the MIT license. The model
implementation and ImageNet initialization remain subject to upstream terms.
A paper citation will be added after publication.