Abstract

We present the next generation of MobileNets based on a combination ofcomplementary search techniques as well as a novel architecture design.MobileNetV3 is tuned to mobile phone CPUs through a combination ofhardware-aware network architecture search (NAS) complemented by the NetAdaptalgorithm and then subsequently improved through novel architecture advances.This paper starts the exploration of how automated search algorithms andnetwork design can work together to harness complementary approaches improvingthe overall state of the art. Through this process we create two new MobileNetmodels for release: MobileNetV3-Large and MobileNetV3-Small which are targetedfor high and low resource use cases. These models are then adapted and appliedto the tasks of object detection and semantic segmentation. For the task ofsemantic segmentation (or any dense pixel prediction), we propose a newefficient segmentation decoder Lite Reduced Atrous Spatial Pyramid Pooling(LR-ASPP). We achieve new state of the art results for mobile classification,detection and segmentation. MobileNetV3-Large is 3.2\% more accurate onImageNet classification while reducing latency by 15\% compared to MobileNetV2.MobileNetV3-Small is 4.6\% more accurate while reducing latency by 5\% comparedto MobileNetV2. MobileNetV3-Large detection is 25\% faster at roughly the sameaccuracy as MobileNetV2 on COCO detection. MobileNetV3-Large LR-ASPP is 30\%faster than MobileNetV2 R-ASPP at similar accuracy for Cityscapes segmentation.