LN_HEV_only_segmentation_sweep

A r2unet model for binary image segmentation trained with sliding window approach.

Model Description

• Architecture: r2unet
• Input Channels: 3
• Output Classes: 1
• Base Filters: 32
• Window Size: 128
• Downsample Factor: 1.0

Model-Specific Parameters

• Recurrence (t): 3

Training Configuration

Parameter	Value
Batch Size	16
Learning Rate	4.274719008769108e-06
Weight Decay	8.030299406726313e-05
Epochs	100
Patience	10
Dataset	GleghornLab/Semi-Automated_LN_Segmentation_10_11_2025

Performance Metrics

Metric	Mean	Class 0
Dice	0.2140	0.2140
IoU	0.1207	0.1207
F1	0.2140	0.2140
MCC	0.2953	0.2953
ROC AUC	0.9242	0.9242
PR AUC	0.1705	0.1705

Usage

import numpy as np
from model import MODEL_REGISTRY, SegmentationConfig

# Load model
config = SegmentationConfig.from_pretrained("aholk/LN_HEV_only_segmentation_sweep")
model = MODEL_REGISTRY["r2unet"].from_pretrained("aholk/LN_HEV_only_segmentation_sweep")
model.eval()

# Run inference on a full image with sliding window
image = np.random.rand(2048, 2048, 3).astype(np.float32)  # Your image here
probs = model.predict_full_image(
    image,
    dim=128,
    batch_size=16,
    device="cuda"  # or "cpu"
)
# probs shape: (num_classes, H, W) with values in [0, 1]

# Threshold to get binary masks
masks = (probs > 0.5).astype(np.uint8)

Citation

If you use this model, please cite:

@software{windowz_segmentation,
  title={Multilabel Image Segmentation with Sliding Window U-Net},
  author={Gleghorn Lab},
  year={2025},
  url={https://github.com/GleghornLab/ComputerVision2}
}