AI Paper Atlas

Research map / Detection and segmentation / CNN features

Accelerating Very Deep Convolutional Networks for Classification and Detection

Xiangyu Zhang, Jianhua Zou, Kaiming He, Jian Zhong Sun

TPAMI 2016 · Detection and segmentation › CNN features · approach: Efficient network training and design · 874 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 asymmetric reconstruction (account for approximated previous layers) (ablation)low-rank layer approximation · SPP-10, 3 layers approximated at 4xtop-5 error increase: over 1.0%Improved
add 3d decomposition (spatial plus channel) to asymmetric reconstruction (ablation)low-rank acceleration · SPP-10 ImageNet 4xtop-5 error (1-view): 16.7 -> 14.1Improved
add fine-tuning to asym. (3d)low-rank acceleration · SPP-10 ImageNet 4xtop-5 error (1-view): 14.1 -> 13.8Improved
add whole-model rank selection (ablation)asymmetric solver · SPP-10 ImageNet 4xincrease of top-5 error: 4.32 -> 4.20Improved
add whole-model rank selection (ablation)asymmetric solver · VGG-16 ImageNet 4xincrease of top-5 error: 6.38 -> 3.84Improved
add whole-model rank selection (ablation)asymmetric solver · VGG-16 ImageNet 2xincrease of top-5 error: 0.99 -> 0.28Improved
use nonlinear solution instead of linear solution (ablation)single-layer low-rank approximation · SPP-10 single-layerincrease of top-5 errorImproved
use our method instead of Jaderberg et al. spatial decompositionwhole-model acceleration · SPP-10 ImageNet 4xincrease of top-5 error: 6.0% -> 4.2%Improved
use our asym. (3d) instead of Jaderberg et al.VGG-16 acceleration · VGG-16 ImageNet 5xincrease of top-5 error: 29.7 -> 2.0Improved
use accelerated 3d model instead of same architecture trained from scratch (ablation)decomposed SPP-10 architecture · SPP-10 ImageNet 4xtop-5 error (1-view): 16.9 -> 14.1Improved
increase conv speedup from 3x to 5xaccelerated VGG-16 + Fast R-CNN · PASCAL VOC 2007mAP: 66.9 -> 65.2Hurt
add fine-tuning to asym. (3d)VGG-16 acceleration at 4x · ImageNet valincrease of top-5 error: 0.9 -> 0.3Improved

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