AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Top-Down Learning for Structured Labeling with Convolutional Pseudoprior

Saining Xie, Xun Huang, Zhuowen Tu

ECCV 2016 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis · 41 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 ConvPPFCN-8s · Pascal-Contextmean IU: 37.8 -> 41.0Improved
add ConvPPFCN-32s · Pascal-Contextmean IU: 35.1 -> 37.1Improved
add ConvPPFCN-16s · Pascal-Contextmean IU: 37.6 -> 40.3Improved
add multi-scale context (pool5 + pool4 + pool3 instead of pool5 only) (ablation)ConvPP · Pascal-Contextmean IU: 37.1 -> 41.0Improved
add ConvPPFCN-8s · SIFT Flowmean IU: 39.5 -> 40.7Improved
add ConvPPFCN-16s · SIFT Flowmean IU: 39.1 -> 39.7Improved
use 11x11 donut filter instead of 7x7 (ablation)ConvPP · Pascal-Contextmean IU: 40.3 -> 41.0Improved
use 11x11 donut filter instead of 7x7 (ablation)ConvPP · SIFT Flowmean IU: 40.7 -> 32.4Hurt
jointly fine-tune the bottom-up FCN-8s with ConvPP (ablation)ConvPP-8s · Pascal-Contextmean IU: 41.0 -> 40.89No clear effect
add ConvPP to the network (ablation)NN without ConvPP · OCR (small training set)error rate (%): 15.73 -> 6.49Improved
add ConvPP to the network (ablation)NN without ConvPP · OCR (large training set)error rate (%): 3.69 -> 0.57Improved
add ConvPP to the network (ablation)NN without ConvPP · FAQ sentence labelingerror rate (%): 5.25 -> 1.09Improved

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