Light-Weight RefineNet for Real-Time Semantic Segmentation

TL;DR

This paper presents a compact, real-time capable version of RefineNet for semantic segmentation, reducing model size and computation while maintaining high accuracy, suitable for high-resolution inputs and real-time applications.

Contribution

The authors adapt RefineNet into a more efficient model by identifying and modifying computationally expensive blocks, achieving over twofold reduction in parameters and FLOPs with minimal performance loss.

Findings

Model reduction over twofold with almost unchanged performance

Speed increase from 20 FPS to 55 FPS on 512x512 inputs

Achieved 79.2% mean IoU with only 3.3M parameters

Abstract

We consider an important task of effective and efficient semantic image segmentation. In particular, we adapt a powerful semantic segmentation architecture, called RefineNet, into the more compact one, suitable even for tasks requiring real-time performance on high-resolution inputs. To this end, we identify computationally expensive blocks in the original setup, and propose two modifications aimed to decrease the number of parameters and floating point operations. By doing that, we achieve more than twofold model reduction, while keeping the performance levels almost intact. Our fastest model undergoes a significant speed-up boost from 20 FPS to 55 FPS on a generic GPU card on 512x512 inputs with solid 81.1% mean iou performance on the test set of PASCAL VOC, while our slowest model with 32 FPS (from original 17 FPS) shows 82.7% mean iou on the same dataset. Alternatively, we showcase that our approach is easily mixable with light-weight classification networks: we attain 79.2% mean iou on PASCAL VOC using a model that contains only 3.3M parameters and performs only 9.3B floating point operations.