WRSWorld Research StateContribute
WORLD RESEARCH STATE

MobileNetV3: hardware-aware search and efficient mobile vision integration

Contribution grade

Efficient vision systems · Human-designed research method or resource

Reference score130,000Provisional editorial estimate of this exact contribution; its grade is assessed separately.

What changed

Combined hardware-aware architecture search, NetAdapt and manual architecture changes for mobile CPUs, with a lightweight LR-ASPP dense-prediction decoder.

Scientific scope

Combined hardware-aware architecture search, NetAdapt and manual architecture changes for mobile CPUs, with a lightweight LR-ASPP dense-prediction decoder. Automated architecture search and NetAdapt helped select network designs; human architectural changes and evaluation also contributed. Inherits MobileNetV2, NAS, NetAdapt and squeeze-excitation. Acquired delta is specific joint search/design and decoder system, not first hardware-aware NAS. Small/Large variants are one contribution and measured latency depends on device/runtime.

Evidence

Automated architecture search and NetAdapt helped select network designs; human architectural changes and evaluation also contributed. WRS has not independently rerun these experiments. Automated architecture search and NetAdapt helped select network designs; human architectural changes and evaluation also contributed. Inherits MobileNetV2, NAS, NetAdapt and squeeze-excitation. Acquired delta is specific joint search/design and decoder system, not first hardware-aware NAS. Small/Large variants are one contribution and measured latency depends on device/runtime.

Contributors

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction