Papers › Improved Conditional VRNNs for Video Prediction

Improved Conditional VRNNs for Video Prediction

27 Apr 2019ICCV 2019 10arXiv:1904.12165archive 2025-07-28

Lluis Castrejon, Nicolas Ballas, Aaron Courville

Predicting future frames for a video sequence is a challenging generative modeling task. Promising approaches include probabilistic latent variable models such as the Variational Auto-Encoder. While VAEs can handle uncertainty and model multiple possible future outcomes, they have a tendency to produce blurry predictions. In this work we argue that this is a sign of underfitting. To address this issue, we propose to increase the expressiveness of the latent distributions and to use higher capacity likelihood models. Our approach relies on a hierarchy of latent variables, which defines a family of flexible prior and posterior distributions in order to better model the probability of future sequences. We validate our proposal through a series of ablation experiments and compare our approach to current state-of-the-art latent variable models. Our method performs favorably under several metrics in three different datasets.

PaperPDFConference PDFCode

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

Code

facebookresearch/improved_vrnn mentioned on GitHubpytorchNOASSERTION report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

PredictionVideo GenerationVideo Prediction

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Video Generation BAIR Robot Pushing Hier-VRNN Cond 2 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing Hier-VRNN FVD score 143.4 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing Hier-VRNN LPIPS 0.055±0.03 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing Hier-VRNN Pred 28 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing Hier-VRNN SSIM 0.822±0.06 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing Hier-VRNN Train 10 #16 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L Cond 2 #18 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L FVD score 149.22 #18 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L LPIPS 0.058±0.03 #18 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L Pred 28 #18 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L SSIM 0.829±0.06 #18 of 31 Archive leaderboard report
Video Generation BAIR Robot Pushing VRNN 1L Train 10 #18 of 31 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN Cond. 2 #2 of 5 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN FVD 567.51 #2 of 5 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN LPIPS 0.264 ± .07 #2 of 5 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN Pred 28 #2 of 5 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN SSIM 0.628±0.1 #2 of 5 Archive leaderboard report
Video Prediction Cityscapes 128x128 Hier-VRNN Train 10 #2 of 5 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections