Tiny Video Networks
This addresses the need for faster video processing in autonomous agents, though it appears incremental as it builds on existing architecture learning methods.
The authors tackled the problem of computationally intensive video understanding by proposing Tiny Video Networks, which automatically designs efficient models that achieve competitive performance with runtimes as low as 37 milliseconds on a CPU and 10 milliseconds on a GPU.
Video understanding is a challenging problem with great impact on the abilities of autonomous agents working in the real-world. Yet, solutions so far have been computationally intensive, with the fastest algorithms running for more than half a second per video snippet on powerful GPUs. We propose a novel idea on video architecture learning - Tiny Video Networks - which automatically designs highly efficient models for video understanding. The tiny video models run with competitive performance for as low as 37 milliseconds per video on a CPU and 10 milliseconds on a standard GPU.