AI Paper Atlas

Research map / Video / Video segmentation and tracking

ContextVP: Fully Context-Aware Video Prediction

Wonmin Byeon, Qin Wang, Rupesh Kumar Srivastava, Petros Koumoutsakos

ECCV 2018 · Video › Video segmentation and tracking · approach: Stochastic video prediction · 152 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
add layers: 1 to 3 (uniform blending) (ablation)ContextVP · Human3.6M next-frame prediction (ablation study)PSNR: 38.1 -> 41.2Improved
add a fourth layer with more units (uniform blending) (ablation)ContextVP3 · Human3.6M next-frame prediction (ablation study)PSNR: 41.2 -> 42.3Improved
use weighted blending instead of uniform blending (ablation)ContextVP4-U-big · Human3.6M next-frame prediction (ablation study)PSNR: 42.3 -> 44.8Improved
add directional weight sharing (ablation)ContextVP4-W-big · Human3.6M next-frame prediction (ablation study)PSNR: 44.8 -> 45.2Improved
increase hidden units (small to big) (ablation)ContextVP4-WD · Human3.6M next-frame prediction (ablation study)PSNR: 45.0 -> 45.2Improved
replace 20-layer ConvLSTM with 4-layer ContextVP4-WD-bigConvLSTM20 · Human3.6M next-frame predictionPSNR: 44.1 -> 45.2Improved
replace 20-layer ConvLSTM with 4-layer ContextVP4-WD-bigConvLSTM20 · CalTech Pedestrian next-frame prediction (trained on KITTI)PSNR: 28.0 -> 28.7Improved
replace 20-layer ConvLSTM with 4-layer ContextVP4-WD-bigConvLSTM20 · UCF-101 next-frame prediction (4 input frames)PSNR: 32.9 -> 34.9Improved
use ContextVP4-WD-big instead of Deep Voxel FlowDVF · UCF-101 next-frame prediction (4 input frames)SSIM: 0.94 -> 0.92Hurt
use 5x5 kernels, nonlinear activations or layer norm in blending blocks (ablation)ContextVP · Human3.6M next-frame predictionPSNRNo clear effect
use L2 loss instead of L1 with GDL (ablation)ContextVP · Human3.6M next-frame predictionPSNRNo clear effect

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