Deepweed v1.1.0 (Bioinformatics revision R1)

Revision snapshot for BIOINF-2026-0506 (major revision R1).

Includes:

• 512x512 retrained petri dish segmentation model (best_instance_unet_512.pt) and training script (train_512.py)
• Independent test set (3 frames, 82 fronds, blank-canvas annotation) and evaluation script (evaluate_segmentation.py)
• Tracking error metrics notebook (evaluate_tracking.ipynb): ID switch rate, fragmentation rate, temporal stability
• Supplementary video rendering with post-stitching labels and stage-drift correction, both as a standalone script (render_stitched.py, detect_shifts.py) and integrated in the analysis notebook (Section 3.1d)

This tag is the code state accompanying the R1 response letter. Archive this release on Zenodo to mint the DOI referenced in the availability statement.

Deepweed v1.0.0

Deepweed v1.0.0

What’s included

• End-to-end duckweed segmentation + tracking + lineage inference pipeline
• Petri dish mode (JPEG/PNG sequence) and microfluidics mode (AVI)
• Pretrained U-Net weights included via Git LFS

How to run

• Follow README installation steps
• Run: paper_analysis/Duckweed_Paper_Analysis.ipynb