Fully Convolutional Networks for Semantic Segmentation

TLDR

Convolutional networks are powerful visual models that yield hierarchies of features. The authors aim to build fully convolutional networks that accept arbitrary‑size inputs and produce correspondingly‑sized outputs with efficient inference and learning. They extend existing classification architectures into fully convolutional forms, fine‑tune them for segmentation, and introduce a skip‑connection design that fuses coarse semantic and fine appearance cues for accurate dense predictions. End‑to‑end fully convolutional networks outperform prior state‑of‑the‑art semantic segmentation, achieving a 30 % relative gain to 67.2 % mean IU on PASCAL VOC 2012 and superior results on NYUDv2, SIFT Flow, and PASCAL‑Context, with inference in under a tenth of a second per image.

Abstract

Convolutional networks are powerful visual models that yield hierarchies of features. We show that convolutional networks by themselves, trained end-to-end, pixels-to-pixels, improve on the previous best result in semantic segmentation. Our key insight is to build "fully convolutional" networks that take input of arbitrary size and produce correspondingly-sized output with efficient inference and learning. We define and detail the space of fully convolutional networks, explain their application to spatially dense prediction tasks, and draw connections to prior models. We adapt contemporary classification networks (AlexNet, the VGG net, and GoogLeNet) into fully convolutional networks and transfer their learned representations by fine-tuning to the segmentation task. We then define a skip architecture that combines semantic information from a deep, coarse layer with appearance information from a shallow, fine layer to produce accurate and detailed segmentations. Our fully convolutional networks achieve improved segmentation of PASCAL VOC (30% relative improvement to 67.2% mean IU on 2012), NYUDv2, SIFT Flow, and PASCAL-Context, while inference takes one tenth of a second for a typical image.