Image Classification - MobileNetV2

This example demonstrates how to perform image classification using a BPU-deployed MobileNetV2 model with inference executed via hbm_runtime. The example code is located in the /app/pydev_demo/01_classification_sample/02_mobilenetv2/ directory.

Model Description

• Overview:

  MobileNetV2 is a lightweight convolutional neural network proposed by Google in 2018, designed for efficient image recognition on mobile devices. It introduces the Inverted Residual and Linear Bottleneck structures to reduce computational cost while improving performance. MobileNetV2 is highly suitable for deployment on edge devices and in resource-constrained scenarios for tasks such as image classification and detection. The MobileNetV2 model used in this example accepts 224×224 input and is a BPU quantized model supporting the NV12 format.

• HBM Model Name: mobilenetv2_224x224_nv12.hbm

• Input Format: NV12, with resolution 224x224 (separated Y and UV planes)

• Output: Softmax probability distribution over 1000 classes (conforming to the ImageNet 1000-class standard)

• Model Download URL (automatically downloaded by the program):

  https://archive.d-robotics.cc/downloads/rdk_model_zoo/rdk_s100/MobileNet/mobilenetv2_224x224_nv12.hbm

Functionality Description

• Model Loading

  Load the model file using hbm_runtime, and extract the model name, input/output names, and their corresponding shapes.

• Input Preprocessing

  Resize the input image from BGR format to 224x224, then convert it to the hardware-required NV12 format (with separated Y and UV planes). The processed data is structured as a dictionary to match the inference interface.

• Inference Execution

  Call the .run() method to perform inference, supporting configuration of BPU core(s) (e.g., core0/core1) and inference priority (0–255).

• Result Post-processing

  Retrieve the model output tensor, parse the softmax probabilities, and display the top-K predictions (default: top-5), including class names and corresponding probabilities.

Environment Dependencies

This example has no special environment requirements—only the dependencies from pydev need to be installed:

pip install -r ../../requirements.txt

Directory Structure

.
├── mobilenetv2.py              # Main inference script
└── README.md                   # Usage instructions

Parameter Description

Parameter	Description	Default Value
--model-path	Path to the model file (.hbm format)	/opt/hobot/model/s100/basic/mobilenetv2_224x224_nv12.hbm
--test-img	Path to the test image	/app/res/assets/zebra_cls.jpg
--label-file	Path to the class label mapping file	/app/res/labels/imagenet1000_clsidx_to_labels.txt
--priority	Model inference priority (0–255; higher value = higher priority)	0
--bpu-cores	List of BPU core IDs to use for inference (e.g., --bpu-cores 0 1)	[0]

Quick Start

• Run the Model
  • With default parameters:

    python mobilenetv2.py

  • With custom parameters:

    python mobilenetv2.py \
    --model-path /opt/hobot/model/s100/basic/mobilenetv2_224x224_nv12.hbm \
    --test-img /app/res/assets/zebra_cls.jpg \
    --label-file /app/res/labels/imagenet1000_clsidx_to_labels.txt

• View Results

  Top-5 Predictions:
  zebra: 0.8916
  tiger, Panthera tigris: 0.0028
  hartebeest: 0.0018
  jaguar, panther, Panthera onca, Felis onca: 0.0016
  tiger cat: 0.0016

Notes

• If the specified model path does not exist, the program will attempt to download the model automatically.