---
license: mit
pipeline_tag: image-segmentation
tags:
  - instance-segmentation
  - maskformer
  - germination
---
# MaskFormer-Germination
Fine-tuned MaskFormer for germination instance segmentation.

## Details
- **Base Model**: `facebook/maskformer-swin-tiny-coco`
- **Classes**: 
  - `0`: Background
  - `1`: Normal
  - `2`: Abnormal
- **Training Data**: 18 images, 31+ annotations per image
- **Epochs**: 5
- **Final Loss**: 1.655
- **Batch Size**: 2
- **Learning Rate**: 5e-5
- **Steps**: 45
- **Runtime**: ~26 minutes

## Usage
```python
from transformers import MaskFormerForInstanceSegmentation, MaskFormerImageProcessor
import torch
from PIL import Image

processor = MaskFormerImageProcessor.from_pretrained("Dreamy0/GermiNet-instance-segmentation-maskformer")
model = MaskFormerForInstanceSegmentation.from_pretrained("Dreamy0/GermiNet-instance-segmentation-maskformer")
model.eval()

image = Image.open("path/to/image.jpg")
inputs = processor(images=image, return_tensors="pt")
with torch.no_grad():
    outputs = model(**inputs)
    results = processor.post_process_instance_segmentation(outputs, target_sizes=[(image.height, image.width)])[0]
    for score, label, mask in zip(results["scores"], results["labels"], results["masks"]):
        if score > 0.5 and label in [1, 2]:
            print(f"Label: {label} ({model.config.id2label[label]}), Score: {score:.3f}, Mask shape: {mask.shape}")