Flow Based Self-supervised Pixel Embedding for Image Segmentation

TL;DR

This paper introduces a self-supervised method for image segmentation that learns features from motion cues by training optical flow estimators on synthetic data and then extracting features from real motion data, improving few-shot segmentation performance.

Contribution

It presents a novel self-supervised framework combining synthetic flow training and motion-based feature learning for image segmentation.

Findings

Significantly improves few-shot segmentation accuracy.

Effective use of synthetic data for optical flow training.

Learned features outperform from-scratch training methods.

Abstract

We propose a new self-supervised approach to image feature learning from motion cue. This new approach leverages recent advances in deep learning in two directions: 1) the success of training deep neural network in estimating optical flow in real data using synthetic flow data; and 2) emerging work in learning image features from motion cues, such as optical flow. Building on these, we demonstrate that image features can be learned in self-supervision by first training an optical flow estimator with synthetic flow data, and then learning image features from the estimated flows in real motion data. We demonstrate and evaluate this approach on an image segmentation task. Using the learned image feature representation, the network performs significantly better than the ones trained from scratch in few-shot segmentation tasks.