AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Deep Transferring Quantization

Zheng Xie, Zhiquan Wen, Jing Liu, Zhiqiang Liu, Xixian Wu, Mingkui Tan

ECCV 2020 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 18 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 AFA loss (attentive feature alignment) to CE (ablation)DTQ (5-bit MobileNetV2) · Caltech 256-30Top-1 accuracy (%): 74.9 -> 79.7Improved
add KL loss on top of CE + AFA (ablation)DTQ (5-bit MobileNetV2) · Caltech 256-30Top-1 accuracy (%): 79.7 -> 80.1Improved
add attentive transfer module (attention) to feature alignment (ablation)DTQ (4-bit MobileNetV2) · Food-101Top-1 accuracy (%): 77.1 -> 78.4Improved
add attentive transfer module (attention) to feature alignment (ablation)DTQ (4-bit MobileNetV2) · Stanford Dogs 120Top-1 accuracy (%): 75.2 -> 76.1Improved
add attentive transfer module (attention) to feature alignment (ablation)DTQ (4-bit MobileNetV2) · CUB-200-2011Top-1 accuracy (%): 71.6 -> 72.1Improved
transfer and quantize simultaneously (one-stage DTQ) instead of two-stage DT then L2-Q (ablation)DTQ (4-bit MobileNetV2) · Stanford Dogs 120Top-1 accuracy (%): 71.5 -> 76.1Improved
add a transfer-only first stage before DTQ (two-stage DT->DTQ) (ablation)DTQ (4-bit MobileNetV2) · Food-101Top-1 accuracy (%): 78.4 -> 79.9Improved
add a transfer-only first stage before DTQ (two-stage DT->DTQ) (ablation)DTQ (4-bit MobileNetV2) · Caltech 256-60Top-1 accuracy (%): 79.9 -> 80.9Improved
use DTQ instead of DELTA-Q (learnable rather than fixed channel attention, plus KL)MobileNetV2 4-bit · Caltech 256-30Top-1 accuracy (%): 74.9 -> 75.9Improved
use DTQ instead of DELTA-QResNet-50 5-bit · CUB-200-2011Top-1 accuracy (%): 80.9 -> 80.3Hurt
use DTQ instead of DELTA-QResNet-50 4-bit · CUB-200-2011Top-1 accuracy (%): 76.4 -> 77.8Improved
use DTQ instead of DELTA-Q4-bit MobileNetV2 with 10 training samples per class · Caltech 256-10Top-1 accuracy (%): +2.6%Improved

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