4.1.2 RDK X3 Model Zoo Usage Guide

Branch and System Requirements

RDK X3 uses the rdk_x3 branch. Recommended system version: RDK OS >= 3.0.0.

git clone https://github.com/D-Robotics/rdk_model_zoo.git
cd rdk_model_zoo
git checkout rdk_x3

Repository Directory Structure

The rdk_x3 branch is organized by demo directory. Main structure:

rdk_model_zoo/
|-- demos/
|   |-- classification/          # Image classification
|   |   |-- GoogLeNet/
|   |   |-- MobileNetV1/
|   |   |-- MobileNetV2/
|   |   |-- MobileNetV4/
|   |   |-- MobileOne/
|   |   |-- RepGhost/
|   |   |-- RepVGG/
|   |   |-- RepViT/
|   |   `-- ResNet/
|   |-- detect/                  # Object detection
|   |   |-- FCOS/
|   |   |-- PaddleOCR/
|   |   |-- YOLOv10/
|   |   |-- YOLOv5/
|   |   `-- YOLOv8/
|   `-- Instance_Segmentation/   # Instance segmentation
|       `-- YOLOv8-Seg/
`-- resource/                    # Documentation resources

Python Inference Interface

The RDK X3 branch provides two Python inference interfaces:

• bpu_infer_lib_x3: A Python inference library that requires manual installation
• hobot_dnn.pyeasy_dnn: A board-builtin Python inference interface

Which interface to use depends on the target demo directory README and source entry point.

caution

bpu_infer_lib_x3 and hobot_dnn.pyeasy_dnn have poor support for featuremap input models. If you need to run featuremap input models, consider using C++ interfaces or upgrading to the RDK X5 platform with the hbm_runtime inference interface.

Installing bpu_infer_lib_x3

For RDK X3 demos that use bpu_infer_lib_x3, install with:

wget -nc https://archive.d-robotics.cc/downloads/rdk_model_zoo/rdk_x3/bpu_infer_lib_x3-1.0.3-py3-none-any.whl
pip install bpu_infer_lib_x3-1.0.3-py3-none-any.whl

Demos that use hobot_dnn.pyeasy_dnn use the board-builtin interface directly, no additional installation required.

Quick Start

Running a Demo

cd demos/<task>/<demo>
less README.md
# Follow the README to prepare models, install dependencies, and run inference.

Task Directory Reference

Task	Directory	Included Models
Image classification	demos/classification	GoogLeNet, MobileNetV1, MobileNetV2, MobileNetV4, MobileOne, RepGhost, RepVGG, RepViT, ResNet
Object detection	demos/detect	FCOS, PaddleOCR, YOLOv10, YOLOv5, YOLOv8
Instance segmentation	demos/Instance_Segmentation	YOLOv8-Seg

Using Jupyter

Some RDK X3 demos provide Jupyter Notebooks for interactive execution:

# Install Jupyter Lab
pip install jupyterlab

# Start Jupyter Lab (replace IP with the actual board IP)
jupyter lab --allow-root --ip 192.168.1.10

After launching, open the link in a browser, navigate to the desired demo's notebook, and click Run to try it out.

note

All relative paths in programs are calculated from the model's directory.