AI Paper Atlas

Research map / Architectures and efficiency / Quantization

FATNN: Fast and Accurate Ternary Neural Networks

Peng Chen, Bohan Zhuang, Chunhua Shen

ICCV 2021 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 26 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 non-uniform step size quantization (ablation)ternary LSQ baseline (ResNet-18) · ImageNet valTop-1 accuracy: 64.7 -> 65.4Improved
add distribution calibration of the residual branch (ablation)FATNN ternary (ResNet-18) · ImageNet valTop-1 accuracy: 65.4 -> 66.0Improved
add distribution calibration of the residual branch (ablation)full-precision ResNet-18 · ImageNet valTop-1 accuracy: 69.8 -> 69.6Hurt
replace LSQ with FATNN quantization (ternary)ResNet-18 · ImageNet valTop-1 accuracy: 64.7 -> 66.0Improved
replace LSQ with FATNN quantization (ternary)ResNet-34 · ImageNet valTop-1 accuracy: 69.0 -> 69.8Improved
replace LSQ with FATNN quantization (ternary)ResNet-50 · ImageNet valTop-1 accuracy: 71.2 -> 72.4Improved
use FATNN ternary instead of LQ-Net 2-bit activation/binary weightResNet-18 · ImageNet valTop-1 accuracy: 62.6 -> 66.0Improved
use FATNN ternary implementation instead of 2-bit implementationResNet-18 on Qualcomm 835 · execution timespeedup: 2.1Improved
use FATNN ternary implementation instead of 2-bit implementationResNet-34 on Qualcomm 821 · execution timespeedup: 2.0Improved
include extra bit operations needed by FATNNternary conv layers on Qualcomm 821 · layer-wise execution timerelative overhead: 0.2% to 10.3%No clear effect
include extra bit operations needed by FATNNternary conv layers on Qualcomm 835 · layer-wise execution timerelative overhead: 0.2% to 32.0%Mixed
include extra bit operations needed by FATNNternary conv layers on Nvidia 2080Ti · layer-wise execution timerelative overhead: 1.7% to 10.3%No clear effect

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