DeepCache: Principled Cache for Mobile Deep Vision
This addresses the challenge of mobile vision systems needing to operate efficiently under video scene variations, offering a deployable solution for off-the-shelf devices.
The paper tackles the problem of improving efficiency in deep learning inference for continuous mobile vision by introducing DeepCache, a cache design that exploits temporal locality in video streams, resulting in an average 18% reduction in inference execution time and up to 47% savings, along with a 20% average reduction in energy consumption.
We present DeepCache, a principled cache design for deep learning inference in continuous mobile vision. DeepCache benefits model execution efficiency by exploiting temporal locality in input video streams. It addresses a key challenge raised by mobile vision: the cache must operate under video scene variation, while trading off among cacheability, overhead, and loss in model accuracy. At the input of a model, DeepCache discovers video temporal locality by exploiting the video's internal structure, for which it borrows proven heuristics from video compression; into the model, DeepCache propagates regions of reusable results by exploiting the model's internal structure. Notably, DeepCache eschews applying video heuristics to model internals which are not pixels but high-dimensional, difficult-to-interpret data. Our implementation of DeepCache works with unmodified deep learning models, requires zero developer's manual effort, and is therefore immediately deployable on off-the-shelf mobile devices. Our experiments show that DeepCache saves inference execution time by 18% on average and up to 47%. DeepCache reduces system energy consumption by 20% on average.