MEDIA ANALYTICS
PROJECTS
PEOPLE
PUBLICATIONS
PATENTS
Media Analytics
Overview: While autonomous cars are rapidly becoming a reality, it remains a challenge to scalably deploy them across geographies and conditions. Our full-stack autonomy solutions include perception, prediction, planning, simulation and DevOps that leverage the latest advances in generative AI, neural rendering, large language models, diffusion models and transformers.
Overview: Multi-task learning commonly encounters competition for resources among tasks when model capacity is limited. We develop neural architectures that allow control over the relative importance of tasks and total compute cost during inference time.
Overview: We develop open vocabulary perception methods that combine the power of vision and language to provide rich descriptions of objects in scenes, including their attributes, behaviors, relations and interactions.
This report aims to provide a comprehensive view on the inference efficiency of DETR-style detection models. We provide the effect of the basic efficiency techniques and identify the factors that are easily applicable yet effectively improve the efficiency-accuracy trade-off. Specifically, we explore
Vision transformer based models bring significant improvements for image segmentation tasks. Although these architectures offer powerful capabilities irrespective of specific segmentation tasks, their use of computational resources can be taxing on deployed devices. One way to overcome this challenge
A powerful architecture for universal segmentation relies on transformers that encode multi-scale image features and decode object queries into mask predictions. With efficiency being a high priority for scaling such models, we observed that the state-of-the-art method Mask2Former uses >50% of its compute
We aim to train a multi-task model such that users can adjust the desired compute budget and relative importance of task performances after deployment, without retraining. This enables optimizing performance for dynamically varying user needs, without heavy computational overhead to train and save models
