MobileNetV2 Int8

MobileNet is a lightweight deep neural network family designed by Google for mobile and embedded devices. By using efficient convolution designs, it significantly reduces parameter count and compute cost, enabling real-time vision workloads on resource-constrained devices such as smartphones and IoT terminals.

• Key features: efficient image classification, object detection, and semantic segmentation with low latency.
• Variant: this guide uses MobileNetV2 Int8, which balances accuracy and efficiency and is well-suited for real-time edge deployments.

Environment setup

Make sure the required environment is ready:

• Environment setup
• AI Model Hub

Quick Start

Download the model files

O6 / O6N

cd ai_model_hub_25_Q3/models/ComputeVision/Image_Classification/onnx_mobilenet_v2_12_int8/model
wget https://modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/ComputeVision/Image_Classification/onnx_mobilenet_v2_12_int8/model/mobilenetv2-12-int8-fix.onnx

Test the model

info

Activate your virtual environment before running.

O6 / O6N

python3 inference_onnx.py --EP NPU

Full Conversion Workflow

Download the model files

Linux PC

cd ai_model_hub_25_Q3/models/ComputeVision/Image_Classification/onnx_mobilenet_v2_12_int8/model
wget https://modelscope.cn/models/cix/ai_model_hub_25_Q3/resolve/master/models/ComputeVision/Image_Classification/onnx_mobilenet_v2_12_int8/model/mobilenetv2-12-int8.onnx

Project structure

├── inference_onnx.py
├── model
├── ReadMe.md
└── test_data

Fix the model input shape

Linux PC

python3 -m onnxruntime.tools.make_dynamic_shape_fixed --dim_param batch_size --dim_value 1 mobilenetv2-12-int8.onnx mobilenetv2-12-int8-fix.onnx

Copy to the device

After the conversion, copy the cix model files to the device.

Test inference on the host

Linux PC

python3 inference_onnx.py --EP CPU

Deploy on the NPU

Export environment variables

O6 / O6N

export LD_LIBRARY_PATH=/usr/share/cix/lib/onnxruntime:$LD_LIBRARY_PATH
export OPERATOR_PATH=/usr/share/cix/lib/onnxruntime/operator/

Run the inference script

O6 / O6N

python3 inference_onnx.py --EP NPU

Inference result

O6 / O6N

$ python3 ./inference_onnx.py --EP npu
image path : ./test_data/ILSVRC2012_val_00037133.JPEG
ice bear, polar bear, Ursus Maritimus, Thalarctos maritimus
image path : ./test_data/ILSVRC2012_val_00021564.JPEG
coucal
image path : ./test_data/ILSVRC2012_val_00024154.JPEG
Ibizan hound, Ibizan Podenco
image path : ./test_data/ILSVRC2012_val_00002899.JPEG
rock python, rock snake, Python sebae
image path : ./test_data/ILSVRC2012_val_00045790.JPEG
Yorkshire terrier