Hardware Guide
For people counting, the STM32H7 with CMSIS-NN scores Excellent. Its 1024 KB internal SRAM (5.3x the required 192 KB) and 480 MHz clock ensure smooth real-time inference on 200 KB models. Hardware DSP extensions boost throughput.
| Spec | STM32H7 |
|---|---|
| Processor | ARM Cortex-M7 @ 480 MHz |
| SRAM | 1024 KB |
| Flash | 2 MB |
| Key Features | Double-precision FPU, L1 cache (16 KB I + 16 KB D), JPEG codec, Chrom-ART Accelerator (DMA2D) |
| Connectivity | Ethernet, USB OTG HS/FS |
| Price Range | $8 - $20 (chip), $30 - $80 (dev board) |
The STM32H7's 1024 KB SRAM provides 5.3x the 192 KB minimum for people counting. This generous headroom means the 200 KB model tensor arena, sensor input buffers, and application logic (camera polling, Ethernet stack, state management) all fit without contention. The remaining 524 KB after model allocation supports complex application features. Flash storage at 2 MB accommodates the CMSIS-NN runtime and 200 KB model. Space remains for firmware and basic OTA capability. The STM32H7 at 480 MHz with double-precision FPU and ART accelerator is among the highest-performance Cortex-M MCUs in ST's lineup. Its 1 MB SRAM accommodates models that smaller MCUs cannot fit in memory. Dual-bank flash enables safe OTA firmware updates during operation. For people counting, connect a camera module (e.g., OV2640 via DVP/SPI) via SPI to the STM32H7. The camera interface supports QVGA (320×240) or lower resolution for on-device inference. Downsample to the model's input size (typically 96×96 or 128×128 pixels) before feeding the neural network. CMSIS-NN provides ARM-optimized neural network kernels that leverage the STM32H7's DSP instructions and floating-point unit for maximum inference throughput on Cortex-M. The kernels are hand-optimized in assembly for critical operations (Conv2D, DepthwiseConv2D, FullyConnected). Combine with TFLite Micro's CMSIS-NN delegate for the best performance on ARM targets. At $8-20 per chip ($30-80 for dev boards), the STM32H7 offers strong value for people counting deployments. 22 PlatformIO-listed boards provide decent hardware selection. Key STM32H7 features for this workload: Double-precision FPU, L1 cache (16 KB I + 16 KB D), JPEG codec, Chrom-ART Accelerator (DMA2D).
Set up STM32H7 development environment
Install STM32CubeIDE with the latest STM32Cube firmware package. Create a project targeting the STM32H7 and verify basic functionality (blink LED, serial output). For CMSIS-NN, clone the framework repository and add it as a library dependency. Ensure the toolchain supports C++11 or later for the ML runtime.
Collect camera training data
Connect a camera module (e.g., OV2640 via DVP/SPI) to the STM32H7. Write a data logging sketch that captures camera readings at the target sample rate and outputs via serial/SD card. Collect 1000+ labeled samples across all classes. Capture images at the model input resolution (96×96 or lower).
Train model and prepare for CMSIS-NN deployment
Train a quantized MobileNet-SSD or YOLO-Tiny in TensorFlow/Keras. Apply int8 post-training quantization via the TFLite converter — this is essential for CMSIS-NN's optimized kernels. The quantized model should be under 200 KB. Use tflite_micro's CMSIS-NN delegate to automatically route operations to optimized ARM kernels on the STM32H7's cortex-m7 core.
Deploy and validate on STM32H7
Include the CMSIS-NN runtime and compiled model in your STMicroelectronics project. Allocate a tensor arena of 300-500 KB in a static buffer. Run inference on live camera data and compare predictions against your test set. Log results to serial for desktop validation. Measure inference latency and peak RAM usage to verify they meet application requirements.
NXP cortex-m7 at 600 MHz with 1024 KB SRAM. $6-12 per chip. Compared to STM32H7: cheaper. Excellent rated.
STMicroelectronics cortex-m7 at 216 MHz with 512 KB SRAM. $8-15 per chip. Compared to STM32H7: less RAM but lower cost. Good rated.
STMicroelectronics cortex-m33 at 160 MHz with 786 KB SRAM. $6-15 per chip. Compared to STM32H7: cheaper. Good rated.
Connect cameras to on-device inference — design detection workflows visually and compile to optimized firmware.
Get Started Free