EdgeSegNet: A Compact Network for Semantic Segmentation
This work addresses the need for efficient semantic segmentation in low-power edge scenarios, representing an incremental improvement in model compression and speed.
The authors tackled the problem of semantic segmentation on edge devices by introducing EdgeSegNet, a compact deep convolutional neural network that achieves comparable accuracy to larger models while being over 20x smaller and running at ~38.5 FPS on an NVidia Jetson AGX Xavier.
In this study, we introduce EdgeSegNet, a compact deep convolutional neural network for the task of semantic segmentation. A human-machine collaborative design strategy is leveraged to create EdgeSegNet, where principled network design prototyping is coupled with machine-driven design exploration to create networks with customized module-level macroarchitecture and microarchitecture designs tailored for the task. Experimental results showed that EdgeSegNet can achieve semantic segmentation accuracy comparable with much larger and computationally complex networks (>20x} smaller model size than RefineNet) as well as achieving an inference speed of ~38.5 FPS on an NVidia Jetson AGX Xavier. As such, the proposed EdgeSegNet is well-suited for low-power edge scenarios.