Derive prediction probability shape from class_names in save_predictions

[gbeane]

This pull request updates how probability arrays are allocated and stored in the save_predictions method, ensuring that the shape of the probability array is determined by the presence of class_names rather than by inspecting the contents of the probabilities dictionary. This change fixes shape mismatches for empty multi-class predictions and is verified with new targeted tests.

Probability array allocation and shape logic:

• src/jabs/project/project.py: The logic for allocating the prediction_prob array in save_predictions now uses the class_names argument to determine shape, ensuring that even when probabilities is empty, the correct (n_identities, n_frames, n_classes) shape is used for multi-class predictions, and (n_identities, n_frames) for binary predictions.

Testing improvements:

• tests/project/test_project.py: Added new tests to verify that save_predictions allocates and writes probability arrays with the correct shape for both empty and non-empty multi-class predictions, as well as for binary predictions.
• tests/project/test_project.py: Added minimal stand-in objects for pose estimation and classifier to support the new tests.
• tests/project/test_project.py: Added necessary imports for h5py and the MULTICLASS_PREDICTION_KEY constant. [1] [2]

[Copilot]

Pull request overview

This pull request updates Project.save_predictions to allocate the probability array shape based on whether class_names is provided (multi-class) vs omitted (binary), fixing shape mismatches when saving empty multi-class predictions, and adds targeted tests to lock in the expected HDF5 output shapes.

Changes:

• Allocate prediction_prob as (n_identities, n_frames, n_classes) when class_names is provided, otherwise (n_identities, n_frames).
• Add tests covering empty and non-empty multi-class saves plus binary saves, asserting stored HDF5 dataset shapes.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
src/jabs/project/project.py	Changes probability-array allocation logic in save_predictions to derive shape from class_names.
tests/project/test_project.py	Adds regression tests verifying saved probability dataset shapes for multi-class and binary predictions.

---

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

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated no new comments.