End-to-End Video Instance Segmentation with Transformers

TLDR

Video instance segmentation requires simultaneously classifying, segmenting, and tracking object instances in video, and recent approaches rely on complex pipelines to address this challenge. The authors introduce VisTR, a Transformer‑based framework that treats VIS as a direct end‑to‑end parallel sequence decoding problem. VisTR processes a video clip and directly outputs a sequence of masks for each instance, employing a novel instance‑sequence matching and segmentation strategy that learns similarity across the entire sequence, thereby simplifying the pipeline and achieving the fastest inference speed and best single‑model performance on YouTube‑VIS. VisTR demonstrates that a simpler, faster Transformer‑based approach can attain competitive accuracy, outperforming existing single‑model methods on YouTube‑VIS and providing code for reproducibility.

Abstract

Video instance segmentation (VIS) is the task that requires simultaneously classifying, segmenting and tracking object instances of interest in video. Recent methods typically develop sophisticated pipelines to tackle this task. Here, we propose a new video instance segmentation framework built upon Transformers, termed VisTR, which views the VIS task as a direct end-to-end parallel sequence decoding/prediction problem. Given a video clip consisting of multiple image frames as input, VisTR outputs the sequence of masks for each instance in the video in order directly. At the core is a new, effective instance sequence matching and segmentation strategy, which supervises and segments instances at the sequence level as a whole. VisTR frames the instance segmentation and tracking in the same perspective of similarity learning, thus considerably simplifying the overall pipeline and is significantly different from existing approaches.Without bells and whistles, VisTR achieves the highest speed among all existing VIS models, and achieves the best result among methods using single model on the YouTube-VIS dataset. For the first time, we demonstrate a much simpler and faster video instance segmentation framework built upon Transformers, achieving competitive accuracy. We hope that VisTR can motivate future research for more video understanding tasks.Code is available at: https://git.io/VisTR