Developing Robust Multi-Task Models for AV Perception
, Senior Deep Learning Engineer, NVIDIA
, Senior Deep Learning Software Engineer, NVIDIA
Multi-task learning is becoming increasingly popular in real-world applications such as autonomous driving. Relevant tasks, when trained together, may provide mutually complementary information that improves the performance of each task. However, challenges in practice, such as data imbalance and domain gaps, may hinder the learning process and result in sub-optimal performance. This session proposes an end-to-end workflow for effective learning of multi-task models under those limitations and sheds light on how the model can be efficiently deployed.