Daisy-Chain Detection & Classification

gst-ai-daisychain-detection-classification performs cascaded inference: first Object Detection locates objects, then Image Classification identifies each detected object for finer-grained recognition.

Prerequisites

• Completed QIM SDK Installation and Model Download

Steps

1. Verify Models

ls -l /etc/models/yolox_quantized.tflite /etc/models/inception_v3_quantized.tflite

2. View Configuration

cat /etc/configs/config_daisychain_detection_classification.json

3. Run

radxa@airbox$

gst-ai-daisychain-detection-classification --config-file=/etc/configs/config_daisychain_detection_classification.json

Press Ctrl + C to stop.

Expected Output

Using DSP delegate with TFLITE for Classification
Using DSP delegate with TFLITE for Detection
VERBOSE: Replacing 329 out of 329 node(s) with delegate (TfLiteQnnDelegate) node
VERBOSE: Replacing 142 out of 142 node(s) with delegate (TfLiteQnnDelegate) node
Pipeline state changed from PAUSED to PLAYING

The display shows the test video with bounding boxes from detection and fine-grained classification labels for each detected object.

Validation

• Detection (YOLOX, 329 ops) and classification (InceptionV3, 142 ops) running simultaneously on DSP
• Pipeline reaches PLAYING state
• Display shows both bounding boxes and classification labels