CVPR 2021 Best Paper Awards Highlight Efficient AI and Generative Scenes
CVPR 2021 reveals its top paper awards, recognizing advancements in sparse video sampling, Siamese learning, and generative neural fields. The winners showcase a strong trend toward data-efficient and high-fidelity AI systems.
The CVPR 2021 awards committee reveals this year's top achievements in computer vision research, highlighting a strong shift toward data efficiency and advanced scene generation. Michael Niemeyer and Andreas Geiger win the Best Paper award for GIRAFFE, a method that represents scenes as compositional generative neural feature fields to create highly controllable 3D image synthesis.
Best Paper honorable mentions go to Xinlei Chen and Kaiming He for exploring simple Siamese representation learning, and to Yasamin Jafarian and Hyun Soo Park for learning high-fidelity depths of dressed humans through social media dance videos. In the student category, Jennifer J. Sun and colleagues take the top prize for Task Programming, which focuses on learning data-efficient behavior representations.
Student paper honorable mentions recognize three impressive projects that tackle diverse visual challenges. These include ClipBERT for video-and-language learning via sparse sampling, Binary TTC for creating temporal geofences in autonomous navigation, and a real-time high-resolution background matting technique that separates subjects from their environments seamlessly.