AI Paper Atlas

Research map / Video / Video segmentation and tracking

Video Scene Parsing With Predictive Feature Learning

Xiaojie Jin, Xin Li, Huaxin Xiao, Xiaohui Shen, Zhe Lin, Jimei Yang, Yunpeng Chen, Jian Dong, Luoqi Liu, Zequn Jie, Jiashi Feng, Shuicheng Yan

ICCV 2017 · Video › Video segmentation and tracking · approach: Stochastic video prediction · 107 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 FPNet frozen spatiotemporal features to the image parser (ablation)VGG16 IPNet · Cityscapes valmIoU: 63.4 -> 68.6Improved
use PPNet features instead of FPNet features (ablation)VGG16 IPNet · Cityscapes valmIoU: +0.5 (68.6 -> 69.1)Improved
train FPNet unsupervised from random init (VGG11) rather than ImageNet-pretrained VGG16 (ablation)FPNet · Cityscapes valmIoU: 68.4 vs 68.6No clear effect
joint end-to-end training of PPNet and IPNet instead of frozen features (ablation)PEARL VGG16 · Cityscapes valmIoU: 69.1 -> 69.8Improved
replace optical flow input channels with PEARL predictive features (ablation)VGG16 parser · Cityscapes valmIoU: 64.5 -> 69.8Improved
replace optical flow input channels with PEARL predictive features (ablation)Res101 parser · Cityscapes valmIoU: 72.7 -> 74.9Improved
add optical flow channels to input (ablation)VGG16 baseline · Cityscapes valmIoU: 63.4 -> 64.5Improved
add optical flow channels to input (ablation)Res101 baseline · Cityscapes valmIoU: 72.5 -> 72.7No clear effect
add AdapNet (one 1x1 conv + ReLU) (ablation)PEARL VGG16 · Cityscapes valmIoU: +1.1Improved
add AdapNet (one 1x1 conv + ReLU) (ablation)PEARL Res101 · Cityscapes valmIoU: +0.3Improved
deeper AdapNet with more conv layers (ablation)PEARL · Cityscapes valmIoU: marginal improvementsNo clear effect
apply PEARL to baselineVGG16 · Cityscapes valmIoU: 63.4 -> 69.8 (+6.4)Improved

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