Mobilenet V1 on STM32H7 using STMCubeMX.AI (v4.1.0)

The repo contains STMWorkbench projects that aim to fit a Mobilenet v1 into an MCU STM32H7 using the STMCubeMX.AI flow. Note that the selected Mobilenet version is the biggest model that can be fitted on the MCU STM32H7 using the new v4.1.0 STMCubeMX.AI flow. In this updated version, 8-bit pre-trained TFLite models implementations are used.

Current Status

The repo is a collection of STMWorkbench projects containing different implementations of MobilenetV1 networks on the STM32H7 board. Particularly, it contains two Mobilenet v1 implementations (3,160,160, alpha=0.25 and 3,192,192, alpha=0.5) targeting the NUCLEO-H7432ZI board.

Content

• MobileNet_v1_0.25_160_quant: STMWorkbench project for MobileNet v1 3,160,160, alpha=0.25 generated throught X-Cube-AI from 8-bit TFLite pre-trained model.
• MobileNet_v1_0.5_192_quant: STMWorkbench project for MobileNet v1 3,192,192, alpha=0.5 generated throught X-Cube-AI from 8-bit TFLite pre-trained model.

Measured Performances

Model	Million MACs	Million Parameters	Top-1 Accuracy	Top-5 Accuracy	CPU Cycles (MCycles)	MMACs/cycle	Latency @480MHz (s)
MobileNet_v1_0.50_192_quant	110	1.34	60.0	82.2	210	0.52	0.437 (2.28 fps)
MobileNet_v1_0.25_160_quant	21	0.47	43.4	68.5	51	0.42	0.106 (9.41 fps)