# Improving Optical Flow and Stereo Depth Estimation by Leveraging Uncertainty-Based Learning Difficulties

arXiv: https://arxiv.org/abs/2506.00324 · HTML: https://mlsanity.com/paper/2506.00324/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 23 of its 32 references that are in the map.
- First posted on arXiv 31 May 2025.

## Builds on (most cited first, 23 of 23)

- [RAFT: Recurrent All-Pairs Field Transforms for Optical Flow](https://mlsanity.com/paper/2003.12039/) (2020)
- [FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence](https://mlsanity.com/paper/2001.07685/) (2020)
- [FlowNet: Learning Optical Flow with Convolutional Networks](https://mlsanity.com/paper/1504.06852/) (2015)
- [A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimation](https://mlsanity.com/paper/1512.02134/) (2015)
- [RAFT-Stereo: Multilevel Recurrent Field Transforms for Stereo Matching](https://mlsanity.com/paper/2109.07547/) (2021)
- [DVC: An End-to-end Deep Video Compression Framework](https://mlsanity.com/paper/1812.00101/) (2018)
- [FlowFormer: A Transformer Architecture for Optical Flow](https://mlsanity.com/paper/2203.16194/) (2022)
- [Learning to Estimate Hidden Motions with Global Motion Aggregation](https://mlsanity.com/paper/2104.02409/) (2021)
- [UnFlow: Unsupervised Learning of Optical Flow with a Bidirectional Census Loss](https://mlsanity.com/paper/1711.07837/) (2017)
- [Practical Stereo Matching via Cascaded Recurrent Network with Adaptive Correlation](https://mlsanity.com/paper/2203.11483/) (2022)
- [AMT: All-Pairs Multi-Field Transforms for Efficient Frame Interpolation](https://mlsanity.com/paper/2304.09790/) (2023)
- [DynamicStereo: Consistent Dynamic Depth from Stereo Videos](https://mlsanity.com/paper/2305.02296/) (2023)
- [Video Compression through Image Interpolation](https://mlsanity.com/paper/1804.06919/) (2018)
- [What Matters in Unsupervised Optical Flow](https://mlsanity.com/paper/2006.04902/) (2020)
- [MaskFlownet: Asymmetric Feature Matching with Learnable Occlusion Mask](https://mlsanity.com/paper/2003.10955/) (2020)
- [IFRNet: Intermediate Feature Refine Network for Efficient Frame Interpolation](https://mlsanity.com/paper/2205.14620/) (2022)
- [SMURF: Self-Teaching Multi-Frame Unsupervised RAFT with Full-Image Warping](https://mlsanity.com/paper/2105.07014/) (2021)
- [LiteFlowNet3: Resolving Correspondence Ambiguity for More Accurate Optical Flow Estimation](https://mlsanity.com/paper/2007.09319/) (2020)
- [Motion Feature Network: Fixed Motion Filter for Action Recognition](https://mlsanity.com/paper/1807.10037/) (2018)
- [DistractFlow: Improving Optical Flow Estimation via Realistic Distractions and Pseudo-Labeling](https://mlsanity.com/paper/2303.14078/) (2023)
- [Temporal Hockey Action Recognition via Pose and Optical Flows](https://mlsanity.com/paper/1812.09533/) (2018)
- [OCAI: Improving Optical Flow Estimation by Occlusion and Consistency Aware Interpolation](https://mlsanity.com/paper/2403.18092/) (2024)
- [SciFlow: Empowering Lightweight Optical Flow Models with Self-Cleaning Iterations](https://mlsanity.com/paper/2404.08135/) (2024)

## Built on by (most cited first, 1 of 1)

- [Geometry-Aware Sparse Depth Sampling for High-Fidelity RGB-D Depth Completion in Robotic Systems](https://mlsanity.com/paper/2512.08229/) (2025)

## Abstract (arXiv)

Conventional training for optical flow and stereo depth models typically employs a uniform loss function across all pixels. However, this one-size-fits-all approach often overlooks the significant variations in learning difficulty among individual pixels and contextual regions. This paper investigates the uncertainty-based confidence maps which capture these spatially varying learning difficulties and introduces tailored solutions to address them. We first present the Difficulty Balancing (DB) loss, which utilizes an error-based confidence measure to encourage the network to focus more on challenging pixels and regions. Moreover, we identify that some difficult pixels and regions are affected by occlusions, resulting from the inherently ill-posed matching problem in the absence of real correspondences. To address this, we propose the Occlusion Avoiding (OA) loss, designed to guide the network into cycle consistency-based confident regions, where feature matching is more reliable. By combining the DB and OA losses, we effectively manage various types of challenging pixels and regions during training. Experiments on both optical flow and stereo depth tasks consistently demonstrate significant performance improvements when applying our proposed combination of the DB and OA losses.