AI Paper Atlas

Research map / Detection and segmentation / CNN features

Boundary-Aware Feature Propagation for Scene Segmentation

Henghui Ding, Xudong Jiang, Ai Qun Liu, Nadia Magnenat Thalmann, Gang Wang

ICCV 2019 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 287 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 unidirectional acyclic graph (UAG) propagation to dilated FCN (ResNet-50) (ablation)FCN · PASCAL-Context (59 classes)mIoU (%): 41.2 -> 49.8Improved
switch backbone from ResNet-50 to ResNet-101 (ablation)BFP with UAGs · PASCAL-Context (59 classes)mIoU (%): 49.8 -> 50.8Improved
add boundary information to control propagation (ablation)BFP with UAGs (ResNet-101) · PASCAL-Context (59 classes)mIoU (%): 50.8 -> 52.8Improved
add multi-scale testing (ablation)BFP (ResNet-101) · PASCAL-Context (59 classes)mIoU (%): 52.8 -> 53.6Improved
replace DAGs with UAGs (ablation)DAG-RNN style propagation · PASCAL-Context (59 classes), VGG-16mIoU (%): 42.6 -> 43.0No clear effect
replace DAGs with UAGs for speed at 960x720 (ablation)dilated FCN · Inference speed, 960x720inference time (s): 56.97 -> 0.76Improved
replace DAGs with UAGs for speed at 480x360 (ablation)dilated FCN · Inference speed, 480x360inference time (s): 17.92 -> 0.47Improved
filter BFP segmentation scores with domain transform (DT)BFP · PASCAL-Context (59 classes)mIoU (%): +0.7Improved
use BFP instead of best prior method EncNetEncNet · PASCAL-Context (59 classes)mIoU (%): 51.7 -> 53.6Improved
use BFP instead of Dense-DecoderDense-Decoder · CamVidmIoU (%): 70.9 -> 74.1Improved
use BFP (ResNet-101) instead of DenseASPP (DenseNet-161)DenseASPP · Cityscapes testmIoU (%): 80.6 -> 81.4Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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