AI Paper Atlas

Research map / Detection and segmentation / CNN features

Predictive Feature Learning for Future Segmentation Prediction

Zihang Lin, Jiangxin Sun, Jian-Fang Hu, Qizhi Yu, Jian-Huang Lai, Wei-Shi Zheng

ICCV 2021 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 18 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 multi-scale feature fusion in the autoencoder (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 short-term: 47.5 -> 48.7Improved
add multi-scale feature fusion in the autoencoder (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 28.8 -> 30.5Improved
add rescaling blocks in encoder and decoder (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 short-term: 45.7 -> 47.5Improved
add rescaling blocks in encoder and decoder (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 26.4 -> 28.8Improved
add the full autoencoder before prediction (ablation)feature prediction model · Cityscapes future instance segmentationAP50 short-term: 44.2 -> 48.7Improved
add the full autoencoder before prediction (ablation)feature prediction model · Cityscapes future instance segmentationAP50 mid-term: 24.7 -> 30.5Improved
add reconstruction loss in the bidirectional prediction module (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 short-term: 47.9 -> 48.7Improved
add reconstruction loss in the bidirectional prediction module (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 29.4 -> 30.5Improved
predict feature residual between neighboring frames instead of next-frame feature (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 short-term: 47.0 -> 48.7Improved
predict feature residual between neighboring frames instead of next-frame feature (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 28.5 -> 30.5Improved
use predictive feature resolution k=4 instead of k=2 (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 23.5 -> 30.5Improved
use predictive feature resolution k=4 instead of k=5 (ablation)Predictive Feature Autoencoder · Cityscapes future instance segmentationAP50 mid-term: 29.4 -> 30.5Improved

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