AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Designing by Training: Acceleration Neural Network for Fast High-Dimensional Convolution

Longquan Dai, Liang Tang, Yuan Xie, Jinhui Tang

NeurIPS 2018 · Architectures and efficiency › CNN design · approach: Invertible and wide-network analysis · 1 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
use AccNet-designed acceleration on bilateral grid instead of handmade BGbilateral grid acceleration · 2D Gaussian filtering sigma=2MSE: 0.952 -> 0.309Improved
use AccNet-designed acceleration on bilateral grid instead of handmade BGbilateral grid acceleration · 5D Gaussian filtering sigma=2MSE: 1.804 -> 0.853Improved
use AccNet-designed acceleration on permutohedral lattice instead of handmade PLpermutohedral lattice acceleration · 2D Gaussian filtering sigma=2MSE: 0.541 -> 0.273Improved
use AccNet-designed acceleration on permutohedral lattice instead of handmade PLpermutohedral lattice acceleration · 5D Gaussian filtering sigma=2MSE: 1.712 -> 0.528Improved
replace three-layer CNN (cascaded 1-D convolutions) with AccNet (ablation)acceleration network · 5D bilateral grid, bandwidth 5filtering error (MSE): 0.473 -> 0.461Improved
replace three-layer CNN with AccNet (ablation)acceleration network · 2D bilateral grid, bandwidth 5filtering error (MSE): 0.245 -> 0.239Improved
replace three-layer CNN with AccNet (ablation)acceleration network · 5D bilateral grid, bandwidth 5training time: 14h -> 7.6hImproved
replace three-layer CNN with AccNet (ablation)acceleration network · 2D bilateral grid, bandwidth 5training time: 12.5h -> 7.2hImproved
use non-Gaussian (smooth exponential family) prior in bilateral solver via AccNet instead of Barron and Poole Gaussianbilateral solver for stereo post-processing · Stereo matching AllRMS: 8.44 -> 7.79Improved
same, non-Gaussian bilateral solverbilateral solver for stereo post-processing · Stereo matching AllMAE: 2.81 -> 2.13Improved
same, non-Gaussian bilateral solverbilateral solver for stereo post-processing · Stereo matching NoOccbad 1%: 11.33 -> 10.41Mixed
accelerate Gaussian mixture kernels in fully connected CRF directly with AccNetfully connected CRF inference · CRF segmentationrunning time: save 60% of the timeImproved

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