RF-Next: Efficient Receptive Field Search for Convolutional Neural Networks

TL;DR

RF-Next introduces a global-to-local search method to automatically discover optimal receptive field combinations, significantly improving model performance across various tasks without manual design.

Contribution

The paper presents a novel global-to-local search scheme for receptive fields, replacing hand-designed patterns with learned combinations, enhancing model adaptability and performance.

Findings

Improved performance on temporal action segmentation

Enhanced object detection and segmentation results

Effective receptive field search method

Abstract

Temporal/spatial receptive fields of models play an important role in sequential/spatial tasks. Large receptive fields facilitate long-term relations, while small receptive fields help to capture the local details. Existing methods construct models with hand-designed receptive fields in layers. Can we effectively search for receptive field combinations to replace hand-designed patterns? To answer this question, we propose to find better receptive field combinations through a global-to-local search scheme. Our search scheme exploits both global search to find the coarse combinations and local search to get the refined receptive field combinations further. The global search finds possible coarse combinations other than human-designed patterns. On top of the global search, we propose an expectation-guided iterative local search scheme to refine combinations effectively. Our RF-Next models, plugging receptive field search to various models, boost the performance on many tasks, e.g., temporal action segmentation, object detection, instance segmentation, and speech synthesis. The source code is publicly available on http://mmcheng.net/rfnext.