# Unsupervised convolutional neural networks for motion estimation

arXiv: https://arxiv.org/abs/1601.06087 · HTML: https://mlsanity.com/paper/1601.06087/

## Key facts

- Cited by 12 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Jan 2016.

## Built on by (most cited first, 12 of 12)

- [PanoFlow: Learning 360deg Optical Flow for Surrounding Temporal Understanding](https://mlsanity.com/paper/2202.13388/) (2022)
- [ChiTransformer:Towards Reliable Stereo from Cues](https://mlsanity.com/paper/2203.04554/) (2022)
- [Self-supervised AutoFlow](https://mlsanity.com/paper/2212.01762/) (2022)
- [ASFlow: Unsupervised Optical Flow Learning with Adaptive Pyramid Sampling](https://mlsanity.com/paper/2104.03560/) (2021)
- [Learning to Estimate Single-View Volumetric Flow Motions without 3D Supervision](https://mlsanity.com/paper/2302.14470/) (2023)
- [DDCNet: Deep Dilated Convolutional Neural Network for Dense Prediction](https://mlsanity.com/paper/2107.04715/) (2021)
- [Scene-Centric Unsupervised Video Panoptic Segmentation](https://mlsanity.com/paper/2606.04925/) (2026)
- [Self-SuperFlow: Self-supervised Scene Flow Prediction in Stereo Sequences](https://mlsanity.com/paper/2206.15296/) (2022)
- [Learning Omnidirectional Flow in 360-degree Video via Siamese Representation](https://mlsanity.com/paper/2208.03620/) (2022)
- [Optical Flow Estimation in 360^circ Videos: Dataset, Model and Application](https://mlsanity.com/paper/2301.11880/) (2023)
- [MaskFlow: Object-Aware Motion Estimation](https://mlsanity.com/paper/2311.12476/) (2023)
- [Weakly supervised alignment and registration of MR-CT for cervical cancer radiotherapy](https://mlsanity.com/paper/2405.12850/) (2024)

## Abstract (arXiv)

Traditional methods for motion estimation estimate the motion field F between a pair of images as the one that minimizes a predesigned cost function. In this paper, we propose a direct method and train a Convolutional Neural Network (CNN) that when, at test time, is given a pair of images as input it produces a dense motion field F at its output layer. In the absence of large datasets with ground truth motion that would allow classical supervised training, we propose to train the network in an unsupervised manner. The proposed cost function that is optimized during training, is based on the classical optical flow constraint. The latter is differentiable with respect to the motion field and, therefore, allows backpropagation of the error to previous layers of the network. Our method is tested on both synthetic and real image sequences and performs similarly to the state-of-the-art methods.