FasterSeg: Searching for Faster Real-time Semantic Segmentation

TL;DR

FasterSeg is an automatically designed semantic segmentation network that achieves state-of-the-art accuracy with significantly faster inference speed by using neural architecture search with a novel search space and a collaborative teacher-student framework.

Contribution

The paper introduces FasterSeg, a NAS-based framework with a new search space and a decoupled latency regularization, plus a co-searching method for teacher-student networks, advancing real-time segmentation.

Findings

FasterSeg runs over 30% faster than manual methods on Cityscapes.

FasterSeg maintains comparable accuracy to state-of-the-art models.

The co-searching framework improves student network accuracy.

Abstract

We present FasterSeg, an automatically designed semantic segmentation network with not only state-of-the-art performance but also faster speed than current methods. Utilizing neural architecture search (NAS), FasterSeg is discovered from a novel and broader search space integrating multi-resolution branches, that has been recently found to be vital in manually designed segmentation models. To better calibrate the balance between the goals of high accuracy and low latency, we propose a decoupled and fine-grained latency regularization, that effectively overcomes our observed phenomenons that the searched networks are prone to "collapsing" to low-latency yet poor-accuracy models. Moreover, we seamlessly extend FasterSeg to a new collaborative search (co-searching) framework, simultaneously searching for a teacher and a student network in the same single run. The teacher-student distillation further boosts the student model's accuracy. Experiments on popular segmentation benchmarks demonstrate the competency of FasterSeg. For example, FasterSeg can run over 30% faster than the closest manually designed competitor on Cityscapes, while maintaining comparable accuracy.