AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Unsupervised Optical Flow Estimation with Dynamic Timing Representation for Spike Camera

Lujie Xia, Ziluo Ding, Rui Zhao, Jiyuan Zhang, Lei Ma, Zhaofei Yu, Tiejun Huang, Ruiqin Xiong

NeurIPS 2023 · Architectures and efficiency › RNNs and normalization · approach: Dataset condensation and flow estimation

Read the paper

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add dynamic timing representation (dilated conv) to PWC-Net(variant), supervisedPWC-Net(variant) · PHM dt=10mean AEE: 1.326 -> 0.854Improved
add dynamic timing representation to RAFT, supervisedRAFT · PHM dt=20mean AEE: 1.904 -> 1.649Improved
replace dilated convolution with normal convolution of same size (ablation)RAFT · PHM dt=10mean AEE: 0.965 -> 1.220Hurt
replace dilated convolution with normal convolutionPWC-Net(variant) · PHM dt=10mean AEE: 1.326 -> 0.932Improved
use dilated convolutions instead of normal convolutions (ablation)PWC-Net(variant) · PHM dt=10mean AEE: 0.932 -> 0.854Improved
add layer attention on top of TMR (ablation)USFlow(pwc) · PHM dt=10 supervisedmean AEE: 0.888 -> 0.854Improved
add layer attention on top of TMR (ablation)USFlow(pwc) · PHM dt=20 supervisedmean AEE: 1.699 -> 1.723Hurt
add layer attention on top of TMR (ablation)USFlow(pwc) · PHM dt=10 unsupervisedmean AEE: 1.355 -> 1.307Improved
add timing representation (TMR) to USFlow(pwc) (ablation)USFlow(pwc) · PHM dt=10 unsupervisedmean AEE: 1.783 -> 1.355Improved
proposed unsupervised loss instead of SSIMUSFlow(pwc) · PHM dt=20 unsupervisedmean AEE: 3.446 -> 2.413Improved
proposed unsupervised loss instead of CensusUSFlow(pwc) · PHM dt=10 unsupervisedmean AEE: 1.759 -> 1.307Improved
remove I_T term from unsupervised loss (ablation)USFlow(pwc) · PHM dt=10 unsupervisedmean AEE: 1.307 -> 2.071Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy