🆕 Add KongNet Model and Example Notebook

[Jiaqi-Lv]

• Add KongNet Model
• Add tests
• Add example notebook (Nucleus Detector)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.41%. Comparing base (a10a84b) to head (4b9eb2f).

Additional details and impacted files

@@                    Coverage Diff                     @@
##           dev-define-engines-abc     #987      +/-   ##
==========================================================
+ Coverage                   95.33%   95.41%   +0.08%     
==========================================================
  Files                          79       80       +1     
  Lines                       10001    10176     +175     
  Branches                     1290     1303      +13     
==========================================================
+ Hits                         9534     9709     +175     
  Misses                        431      431              
  Partials                       36       36              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This pull request adds the KongNet model, a multi-head encoder-decoder architecture for nuclei detection and classification in digital pathology images. The model uses an EfficientNetV2-L encoder and includes 6 pretrained variants for different datasets and challenges.

Changes:

• Added KongNet model implementation with support for multiple decoder heads and attention mechanisms
• Refactored SegmentationHead from grandqc.py to utils.py for reusability
• Updated zarr API calls to use the new zarr_array_kwargs parameter
• Added utility functions for attention mechanisms and NMS post-processing
• Added comprehensive test coverage for the new model and utilities

Reviewed changes

Copilot reviewed 8 out of 9 changed files in this pull request and generated 9 comments.

Show a summary per file

File	Description
tiatoolbox/models/architecture/kongnet.py	New file implementing KongNet multi-head nuclei detection model with TimmEncoder, decoder blocks, and post-processing
tiatoolbox/models/architecture/utils.py	Added SegmentationHead, AttentionModule, SCSEModule classes and nms_on_detection_maps function; enhanced peak_detection_map_overlap with probability output
tiatoolbox/models/architecture/grandqc.py	Refactored to import SegmentationHead from utils instead of local definition
tiatoolbox/models/engine/nucleus_detector.py	Updated zarr API call to use zarr_array_kwargs; improved comment clarity (segmentation→tissue masks)
tiatoolbox/models/engine/engine_abc.py	Updated zarr API call to use zarr_array_kwargs parameter
tiatoolbox/data/pretrained_model.yaml	Added 6 KongNet model configurations for different datasets (CoNIC, PanNuke, MONKEY, PUMA, MIDOG)
tests/models/test_arch_utils.py	Added tests for nms_on_detection_maps and AttentionModule
tests/models/test_arch_kongnet.py	Added comprehensive test suite for KongNet model components and pretrained configurations

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 8 out of 10 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The to_zarr call on line 777 also needs to be updated to use the new API. It should be changed from:

task = dask_output.to_zarr(
    url=save_path, component=key, compute=False, object_codec=None
)

to:

task = dask_output.to_zarr(
    url=save_path, component=key, compute=False, zarr_array_kwargs={"object_codec": None}
)

This is the same API change that was applied to line 790 in this PR.

[Copilot]

The method has a return type annotation of None but is returning the result of super().load_state_dict(). This is inconsistent. Either remove the return keyword (if the return value is not needed), or change the return type annotation to match PyTorch's load_state_dict which returns _IncompatibleKeys. For consistency with the parent class, consider changing the return type annotation to match or simply not returning the value.