Making NeRFs even more compact and efficient using an elastic architecture (and no additional training cost)!
We introduce a generative neural architecture search strategy that generates compact, scene-specialized NeRF architectures.
We introduce a large-scale synthetic food image dataset for nutrition estimation. NV-Synth contains 84,984 photorealistic meal images rendered from 7,082 dynamically plated 3D scenes.
DARLEI combines evolutionary algorithms with parallelized reinforcement learning for efficiently training and evolving populations of UNIMAL agents.
We explore the generation of fast vision transformer architecture designs via generative architecture search to achieve a strong balance between accuracy and architectural and computational efficiency.
We propose MAPLE-Edge, an edge-device oriented latency predictor where we train a regression network on architecture-latency pairs in conjunction with a hardware-runtime descriptor to effectively estimate inference latency on a diverse pool of edge devices.
