It uses a novel pipeline to detect and identify tasks in domain-incremental learning scenarios without supervision, addressing the challenge of data diversity and distribution shift common in autonomous driving data.
TADIL: Task-Agnostic Domain-Incremental Learning through Task-ID Inference using Transformer Nearest-Centroid Embeddings
Climate BSC Group: Computer Sciences SoftwareTADIL (Task-Agnostic Domain-Incremental Learning) is a joint BSC/Lenovo effort in the autonomous driving domain that incorporates Continual Learning concepts to adaptively process data from diverse urban and rural driving environments. TADIL's main objective is to create a domain-incremental detector capable of handling various real-world scenarios like changing weather, time of the day, and locations.
Contact:
License:
Apache License (Version 2.0)
Primary tabs
