# FlowFeat: Pixel-Dense Embedding of Motion Profiles

Source: [FlowFeat: Pixel-Dense Embedding of Motion Profiles](<https://www.sota2.com/research/paper/691414a42828b7cc6937ecc6>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-11-13T05:03:13.274Z

Authors: Nikita Araslanov, Anna Sonnweber, Daniel Cremers

Published: 2025-11-10T23:39:53.000Z

## Abstract

Dense and versatile image representations underpin the success of virtually all computer vision applications. However, state-of-the-art networks, such as transformers, produce low-resolution feature grids, which are suboptimal for dense prediction tasks. To address this limitation, we present FlowFeat, a high-resolution and multi-task feature representation. The key ingredient behind FlowFeat is a novel distillation technique that embeds a distribution of plausible apparent motions, or motion profiles. By leveraging optical flow networks and diverse video data, we develop an effective self-supervised training framework that statistically approximates the apparent motion. With its remarkable level of spatial detail, FlowFeat encodes a compelling degree of geometric and semantic cues while exhibiting high temporal consistency. Empirically, FlowFeat significantly enhances the representational power of five state-of-the-art encoders and alternative upsampling strategies across three dense tasks: video object segmentation, monocular depth estimation and semantic segmentation. Training FlowFeat is computationally inexpensive and robust to inaccurate flow estimation, remaining highly effective even when using unsupervised flow networks. Our work takes a step forward towards reliable and versatile dense image representations.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Semantic Segmentation on COCO Stuff (val)](<https://www.sota2.com/research/sota/semantic-segmentation-on-coco-stuff-val.md>): Semantic segmentation; COCO Stuff (val); mIoU: 60.4.

- [Video Object Segmentation on DAVIS 2017 (val)](<https://www.sota2.com/research/sota/video-object-segmentation-on-davis-2017-val.md>): Video Object Segmentation; DAVIS 2017 (val); J mean: 66.9.

- [Surface Normal Estimation on NYUv2 (val)](<https://www.sota2.com/research/sota/surface-normal-estimation-on-nyuv2-val.md>): Surface Normal Estimation; NYUv2 (val); mAE: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2511.07696>)

[PDF](<https://arxiv.org/pdf/2511.07696.pdf>)
