AI Paper Atlas

Research map / Architectures and efficiency / Quantization

HMQ: Hardware Friendly Mixed Precision Quantization Block for CNNs

Hai Victor Habi, Roy H. Jennings, Arnon Netzer

ECCV 2020 · Architectures and efficiency › Quantization · approach: Mixed-precision quantization · 58 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 HMQ instead of HAQ at ~16x weight compressionMobileNetV1 quantization · ImageNet (WCR ~16)top-1 accuracy (%): 57.14 -> 68.36Improved
use HMQ instead of HAQ at ~16x weight compressionResNet-50 quantization · ImageNet (WCR ~16)top-1 accuracy (%): 70.63 -> 75Improved
use HMQ instead of HAQ at ~16x weight compressionMobileNetV2 quantization · ImageNet (WCR ~16)top-1 accuracy (%): 66.75 -> 65.7Hurt
use HMQ instead of HAQ at ~10x weight compressionMobileNetV1 quantization · ImageNet (WCR ~10)top-1 accuracy (%): 67.66 -> 69.88Improved
use HMQ instead of HAQ at ~10x weight compressionMobileNetV2 quantization · ImageNet (WCR ~10)top-1 accuracy (%): 70.9 -> 70.12Hurt
use HMQ instead of HAQ at ~10x weight compressionResNet-50 quantization · ImageNet (WCR ~10)top-1 accuracy (%): 75.30 -> 76.1Improved
use HMQ instead of HAQ at ~8x weight compressionMobileNetV1 quantization · ImageNet (WCR ~8)top-1 accuracy (%): 71.74 -> 70.912Hurt
use HMQ instead of HAQ at ~8x weight compressionResNet-50 quantization · ImageNet (WCR ~8)top-1 accuracy (%): 76.14 -> 76.3Improved
use HMQ instead of DQ with ACR 8MobileNetV2 quantization · ImageNet (ACR 8)top-1 accuracy (%): 69.74 -> 70.9Improved
use HMQ instead of HAWQ-V2 with ACR 8ResNet-50 quantization · ImageNet (ACR 8)top-1 accuracy (%): 75.7 -> 75.45Mixed
raise target weight compression Rw from 4 to 8HMQ on EfficientNet-B0 · ImageNet (ACR 4)top-1 accuracy (%): 76.4 -> 76Hurt
raise target weight compression Rw from 8 to 12HMQ on EfficientNet-B0 · ImageNet (ACR 4)top-1 accuracy (%): 76 -> 74.6Hurt

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