Learning by Distillation: A Self-Supervised Learning Framework for Optical Flow Estimation

TL;DR

DistillFlow introduces a self-supervised knowledge distillation framework for optical flow estimation, effectively learning from unlabeled data and achieving state-of-the-art results on KITTI and Sintel datasets.

Contribution

The paper proposes a novel self-supervised distillation approach that improves optical flow learning, especially for occluded pixels, and demonstrates superior performance and generalization capabilities.

Findings

Achieves state-of-the-art unsupervised performance on KITTI and Sintel.

Pre-trained models serve as excellent initialization for supervised fine-tuning.

Outperforms all published methods on the Sintel Final benchmark.

Abstract

We present DistillFlow, a knowledge distillation approach to learning optical flow. DistillFlow trains multiple teacher models and a student model, where challenging transformations are applied to the input of the student model to generate hallucinated occlusions as well as less confident predictions. Then, a self-supervised learning framework is constructed: confident predictions from teacher models are served as annotations to guide the student model to learn optical flow for those less confident predictions. The self-supervised learning framework enables us to effectively learn optical flow from unlabeled data, not only for non-occluded pixels, but also for occluded pixels. DistillFlow achieves state-of-the-art unsupervised learning performance on both KITTI and Sintel datasets. Our self-supervised pre-trained model also provides an excellent initialization for supervised fine-tuning, suggesting an alternate training paradigm in contrast to current supervised learning methods that highly rely on pre-training on synthetic data. At the time of writing, our fine-tuned models ranked 1st among all monocular methods on the KITTI 2015 benchmark, and outperform all published methods on the Sintel Final benchmark. More importantly, we demonstrate the generalization capability of DistillFlow in three aspects: framework generalization, correspondence generalization and cross-dataset generalization.