Challenge Details
- Track objects with missing or newly combined sensor modalities.
- Efficient training under limited paired data constraints.
- Calibration-light fusion and modality-agnostic features.
- Evaluation on unreleased, newly collected samples.
Train with partially paired modalities; deploy with full or missing sets using modality-agnostic representations.
.jpg)
.jpg)