AI Paper Atlas

Research map / Video / Video segmentation and tracking

A Hierarchical Variational Neural Uncertainty Model for Stochastic Video Prediction

Moitreya Chatterjee, Narendra Ahuja, Anoop Cherian

ICCV 2021 · Video › Video segmentation and tracking · approach: Stochastic video prediction · 15 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

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
use gamma hyperprior instead of Uniform[0,1] prior (ablation)NUQ · SMMNIST (2,000 training samples)SSIM at convergence: 0.8523 -> 0.8948Improved
add variance encoder-decoder for uncertainty instead of estimating b_t from the decoder (ablation)NUQ · SMMNIST (2,000 training samples)SSIM at convergence: 0.7828 -> 0.8948Improved
add variance encoder-decoder (not deterministic mapping from latent variance) (ablation)NUQ · SMMNIST (2,000 training samples)SSIM at convergence: 0.7648 -> 0.8948Improved
add adversarial discriminator (ablation)NUQ · SMMNIST (2,000 training samples)SSIM at convergence: 0.8929 -> 0.8948No clear effect
add adversarial discriminator (ablation)NUQ · SMMNIST (2,000 training samples)LPIPS at convergence: 0.1967 -> 0.1803Improved
add explicit uncertainty modeling (NUQ) vs Denton and FergusDenton and Fergus · SMMNIST (2,000 training samples)SSIM after 1 epoch: 0.8145 -> 0.8686Improved
add NUQ vs Castrejon et al. with full BAIR training setCastrejon et al. · BAIR Push (43,264)PSNR at convergence: 19.59 -> 19.31Hurt
add NUQ vs Castrejon et al. on KTHCastrejon et al. · KTH ActionSSIM at convergence: 0.8510 -> 0.8448Hurt
add NUQ vs Denton and Fergus with 2,000 training samplesDenton and Fergus · BAIR Push (2,000)PSNR at convergence: 18.49 -> 19.26Improved
add NUQ vs Hsieh et al. on KTHHsieh et al. · KTH ActionSSIM at convergence: 0.7057 -> 0.8448Improved

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