Fully Convolutional Instance-Aware Semantic Segmentation

TLDR

FCNs provide strong semantic segmentation capabilities, while instance mask proposals enable object‑level delineation. This work introduces the first fully convolutional, end‑to‑end solution for instance‑aware semantic segmentation. The architecture jointly predicts instance masks and class labels using a fully shared convolutional backbone across all regions of interest, yielding a highly integrated and efficient network. The method attains state‑of‑the‑art accuracy and efficiency, winning the COCO 2016 segmentation competition by a large margin. Code will be released at https://github.com/daijifeng001/TA-FCN.

Abstract

We present the first fully convolutional end-to-end solution for instance-aware semantic segmentation task. It inherits all the merits of FCNs for semantic segmentation [29] and instance mask proposal [5]. It performs instance mask prediction and classification jointly. The underlying convolutional representation is fully shared between the two sub-tasks, as well as between all regions of interest. The network architecture is highly integrated and efficient. It achieves state-of-the-art performance in both accuracy and efficiency. It wins the COCO 2016 segmentation competition by a large margin. Code would be released at https://github.com/daijifeng001/TA-FCN.