Hardware-Aware Mobile Building Block Evaluation for Computer Vision

In this paper, we propose a methodology to accurately evaluate and compare the performance of efficient neural network building blocks for computer vision in hardware-aware manner. Our comparison uses pareto fronts based on randomly sampled networks from design space capture underlying accuracy/complexity trade-offs. We show that our approach enables matching information obtained by previous paradigms, but provides more insights into relationship between hardware cost accuracy. use analyze different their range embedded platforms. This highlights importance benchmarking as preselection step process network. choosing right block can speed up inference factor two specific ML accelerators.