# Feature Flow: In-network Feature Flow Estimation for Video Object Detection

arXiv: https://arxiv.org/abs/2009.09660 · HTML: https://mlsanity.com/paper/2009.09660/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Sep 2020.

## Built on by (most cited first, 6 of 6)

- [TransVOD: End-to-End Video Object Detection with Spatial-Temporal Transformers](https://mlsanity.com/paper/2201.05047/) (2022)
- [Objects do not disappear: Video object detection by single-frame object location anticipation](https://mlsanity.com/paper/2308.04770/) (2023)
- [Explore Spatio-temporal Aggregation for Insubstantial Object Detection: Benchmark Dataset and Baseline](https://mlsanity.com/paper/2206.11459/) (2022)
- [Object Detection Difficulty: Suppressing Over-aggregation for Faster and Better Video Object Detection](https://mlsanity.com/paper/2308.11327/) (2023)
- [Detector-Augmented SAMURAI for Long-Duration Drone Tracking](https://mlsanity.com/paper/2601.04798/) (2026)
- [Learn Temporal Consistency For Robust Satellite Video Detector](https://mlsanity.com/paper/2606.15112/) (2026)

## Abstract (arXiv)

Optical flow, which expresses pixel displacement, is widely used in many computer vision tasks to provide pixel-level motion information. However, with the remarkable progress of the convolutional neural network, recent state-of-the-art approaches are proposed to solve problems directly on feature-level. Since the displacement of feature vector is not consistent to the pixel displacement, a common approach is to:forward optical flow to a neural network and fine-tune this network on the task dataset. With this method,they expect the fine-tuned network to produce tensors encoding feature-level motion information. In this paper, we rethink this de facto paradigm and analyze its drawbacks in the video object detection task. To mitigate these issues, we propose a novel network (IFF-Net) with an In-network Feature Flow estimation module (IFF module) for video object detection. Without resorting pre-training on any additional dataset, our IFF module is able to directly produce feature flow which indicates the feature displacement. Our IFF module consists of a shallow module, which shares the features with the detection branches. This compact design enables our IFF-Net to accurately detect objects, while maintaining a fast inference speed. Furthermore, we propose a transformation residual loss (TRL) based on textitself-supervision, which further improves the performance of our IFF-Net. Our IFF-Net outperforms existing methods and sets a state-of-the-art performance on ImageNet VID.