v0.24.0

v0.24.0 (2026-05-10)

Bug Fixes

• yolo: Run SAHI eval via preserve-channel slicer always (3a58a54)

Co-authored-by: Cursor cursoragent@cursor.com

Chores

• Ignore plans directory (1d57d8a)

Co-authored-by: Cursor cursoragent@cursor.com

Features

• Unify patch eval JSON for UNet and YOLO (f8a1d1b)

Co-authored-by: Cursor cursoragent@cursor.com

---

Detailed Changes: v0.23.0...v0.24.0

v0.23.0

v0.23.0 (2026-05-10)

Bug Fixes

• Export full test YOLO patches without tile train/val split (6cc47b3)

Co-authored-by: Cursor cursoragent@cursor.com

Chores

• slurm: Reorganize training and test scripts (bb8eae1)

Made-with: Cursor

Features

• Add U-Net patch mask cropper aligned to YOLO exports (c1d33d8)

• Add crop_unet_masks_from_yolo_patches.py, SLURM wrappers, and unit tests\n- Extend evaluation and train_yolo SLURM tests; refresh README; remove CHANGELOG\n- Point train_yolo usage at train_yolo_submit.sh

Co-authored-by: Cursor cursoragent@cursor.com

• slurm: Align pipeline with train_/test_ dataset TIFF naming (59500bd)

Co-authored-by: Cursor cursoragent@cursor.com

• yolo: Add SAHI instance metrics matching U-Net eval (79dfbaa)

• yolo: Export evaluation metrics and artifacts (0e36c77)

Made-with: Cursor

Performance Improvements

• eval: Remove COCO AP and speed up instance metrics (d4e6258)

---

Detailed Changes: v0.22.0...v0.23.0

v0.22.0

v0.22.0 (2026-03-22)

Bug Fixes

• Add missing slurm script (fc42444)

• Added missing dep (again) (87e71e5)

Documentation

• Update readme (d8966b2)

Features

• Add caching (cbbd4d8)

• Per-variant watershed eval and training test fixes (2a0c7d4)

• evaluation: Coco mask AP for U-Net; align AJI instance method (1620efc)

• evaluation: Instance-only metrics and fix PRF empty cases (500f22d)

---

Detailed Changes: v0.21.1...v0.22.0

v0.21.1

v0.21.1 (2026-03-21)

Bug Fixes

• Add missing dependency (0f3c47a)

---

Detailed Changes: v0.21.0...v0.21.1

v0.21.0

v0.21.0 (2026-03-20)

Bug Fixes

• yolo: Val name, manifest paths, SAHI metrics JSON (7d67e4d)

• Always pass --name to Ultralytics val; resolve manifest paths vs manifest dir

• Empty-GT mask AP uses -1 sentinels; aggregate mean_* uses null + strict JSON

• mkdir for nested --output-json; docs and SLURM usage aligned

Features

• Add watershed validation tuning script and SLURM job (bd33972)

• Asdded script to submit watershed tuning for all variants (d4741a8)

• yolo: Run Ultralytics eval on test split, default name test (cf41685)

---

Detailed Changes: v0.20.0...v0.21.0

v0.20.0

v0.20.0 (2026-03-20)

Features

• evaluation: Instance masks from semantic preds (CC + watershed) (3e7a81a)

Refactoring

• yolo: Simplify eval to val/sahi, rename sahi-coco (676bb09)

---

Detailed Changes: v0.19.0...v0.20.0

v0.19.0

v0.19.0 (2026-03-19)

Bug Fixes

• Fixed outdated tests (4d012bd)

Features

• Added yolo evaluation (afd0236)

---

Detailed Changes: v0.18.0...v0.19.0

v0.18.0

v0.18.0 (2026-03-19)

Bug Fixes

• Add missing dep declaration in evaluation package (f97035a)

• Change training time (787134b)

Features

• Add script to make tuning figures (2f2cd33)

---

Detailed Changes: v0.17.0...v0.18.0

v0.17.0

v0.17.0 (2026-03-18)

Features

• Enhance YOLO training and tuning capabilities (70ab526)

• Added hyperparameter tuning options in train_yolo26x_seg.sh and train.py, allowing users to specify tuning epochs and iterations.

• Updated pipeline.py to include a new tune_model function for executing Ultralytics' built-in hyperparameter tuning with a custom search space.

• Modified submit_yolo_experiments.sh to support a new --tune argument for submitting tuning jobs.

• Adjusted README documentation to reflect new tuning features and usage instructions.

• Updated tests to cover new tuning functionality and ensure correct behavior.

• Enhance YOLO training scripts with learning rate and dropout parameters (b5acae5)

• Updated submit_yolo_experiments.sh to accept learning rate and dropout as arguments for job submissions.

• Modified train_yolo26x_seg.sh to include command-line options for learning rate and dropout.

• Adjusted pipeline.py and train.py to utilize the new learning rate and dropout parameters during model training.

• Updated unit tests in test_pipeline.py to validate the new parameters in training and tuning functions.

---

Detailed Changes: v0.16.0...v0.17.0

v0.16.0

v0.16.0 (2026-03-15)

Bug Fixes

• Adjusted batch size and epochs (ef29669)

• Update dataset YAML path handling in YOLO training script (9cc1616)

• Modified train_yolo26x_seg.sh to copy the dataset YAML to a temporary directory and update its path to ensure local resolution of images.

• Added a Python script to rewrite the copied YAML file with the correct dataset path.

• Updated file permissions for submit_yolo_experiments.sh to make it executable.

• Introduced a new test case in test_slurm_scripts.py to verify the YAML path rewriting functionality.

Features

• Add script to split stacked TIFF channels into RGB TIFFs (859d006)

• Introduced split_tiff_channels.py to split a stacked TIFF file into individual RGB TIFF files based on channel triplets.

• Implemented command-line interface for specifying input file, output directory, and optional filename prefix.

• Added validation for input file format and channel count.

• Created unit tests in test_split_tiff_channels.py to ensure correct functionality and error handling.

---

Detailed Changes: v0.15.0...v0.16.0