✨ Feature Description
The framwork is mostly hardcoded for binary classification tasks. It would be very useful to support a multi-class / multi-label classification setting (for lesion classification tasks).
🎯 Problem Statement
All models are implemented with a single-value output. ABMIL can be easily adapted for multiple classes, but for other models, substantial modifications are required.
📌 Use Case
In computational pathology, lots of tasks (regarding lesion detection) require a multi-label/multi-class classification setting.
💡 Proposed Solution
Add support for a "num_classes" parameter in the implemented MIL models to indicate the number of outputs of the model.
✨ Feature Description
The framwork is mostly hardcoded for binary classification tasks. It would be very useful to support a multi-class / multi-label classification setting (for lesion classification tasks).
🎯 Problem Statement
All models are implemented with a single-value output. ABMIL can be easily adapted for multiple classes, but for other models, substantial modifications are required.
📌 Use Case
In computational pathology, lots of tasks (regarding lesion detection) require a multi-label/multi-class classification setting.
💡 Proposed Solution
Add support for a "num_classes" parameter in the implemented MIL models to indicate the number of outputs of the model.