Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells

TL;DR

This paper introduces a fast neural architecture search method for creating compact, high-performance semantic segmentation models that can operate in real-time, using auxiliary cells and efficient search strategies to reduce computational costs.

Contribution

It proposes a novel NAS approach with auxiliary cells and a progressive search strategy, enabling efficient discovery of compact segmentation architectures within limited resources.

Findings

Achieves state-of-the-art performance among compact models in 8 GPU-days.

Uses auxiliary cells for over-parameterization during training.

Demonstrates effectiveness on segmentation, pose estimation, and depth prediction.

Abstract

Automated design of neural network architectures tailored for a specific task is an extremely promising, albeit inherently difficult, avenue to explore. While most results in this domain have been achieved on image classification and language modelling problems, here we concentrate on dense per-pixel tasks, in particular, semantic image segmentation using fully convolutional networks. In contrast to the aforementioned areas, the design choices of a fully convolutional network require several changes, ranging from the sort of operations that need to be used---e.g., dilated convolutions---to a solving of a more difficult optimisation problem. In this work, we are particularly interested in searching for high-performance compact segmentation architectures, able to run in real-time using limited resources. To achieve that, we intentionally over-parameterise the architecture during the training time via a set of auxiliary cells that provide an intermediate supervisory signal and can be omitted during the evaluation phase. The design of the auxiliary cell is emitted by a controller, a neural network with the fixed structure trained using reinforcement learning. More crucially, we demonstrate how to efficiently search for these architectures within limited time and computational budgets. In particular, we rely on a progressive strategy that terminates non-promising architectures from being further trained, and on Polyak averaging coupled with knowledge distillation to speed-up the convergence. Quantitatively, in 8 GPU-days our approach discovers a set of architectures performing on-par with state-of-the-art among compact models on the semantic segmentation, pose estimation and depth prediction tasks. Code will be made available here: https://github.com/drsleep/nas-segm-pytorch