GoogLeNet Example

This document describes using the QAI AppBuilder Python API to run inference with the GoogLeNet image classification model on the Qualcomm® Hexagon™ Processor (NPU).

Supported devices

Device	SoC
Dragon Q6A	QCS6490
Dragon Q8B	SC8280XP
Fogwise® AIRbox Q900	QCS9075

Install QAI AppBuilder

tip

1. Install QAI AppBuilder by following the QAI AppBuilder installation guide.

2. Configure ADSP environment variables as described in Create ADSP environment variables.

Run the sample

Install dependencies

Install sample dependencies in the activated virtual environment:

Device

pip3 install requests tqdm qai-hub py3-wget Pillow torch torchvision opencv-python-headless

Run the script

• Enter the upstream samples directory

  Device

  cd qai-appbuilder/samples

Run inference (Linux requires --chipset; the first run downloads dog.jpg, labels, and the model)

tip

If the auto-downloaded dog.jpg is empty or corrupt, replace it with a valid local image or pass --image /path/to/image.jpg.

QCS6490

Device

python3 ComputerVision/Image_Classification/googlenet/googlenet.py --chipset 6490

SC8280XP

tip

SC8280XP and QCS6490 both use Hexagon V68. Use --chipset 6490 when downloading models.

Device

python3 ComputerVision/Image_Classification/googlenet/googlenet.py --chipset 6490

QCS9075

Device

python3 ComputerVision/Image_Classification/googlenet/googlenet.py --chipset 9075

Expected result

On Dragon Q6A (--chipset 6490), the default test image produces output similar to:

Top 5 predictions for image:

Samoyed 0.9282982349
West Highland White Terrier 0.0102122389
Pomeranian 0.0064557223
Pyrenean Mountain Dog 0.0059806537
Chow Chow 0.0051328284

HTP / FastRPC WARNING lines may appear and can be ignored when inference succeeds.

tip

Launcher alternative:

python3 run_inference.py --model googlenet --args "--chipset 6490"