APLUX Online AI Toolchain

Model Farm

The Model Farm is APLUX's official model resource platform, offering 400+ AI models optimized for Qualcomm platforms, including measured performance benchmarks and ready-to-use sample code.

Core Features

• Query the inference performance of specific models on the A8550MA1 (latency, accuracy, memory usage)
• Download optimized model files adapted for the QNN backend (supports INT4/INT8/FP16 precision)
• Obtain corresponding Python/C++ inference sample code for each model
• Reference model adaptation and optimization steps to migrate custom fine-tuned models

Standard Usage Workflow (using YOLOv5s as an example)

1. Visit the Model Farm: https://aiot.aidlux.com/zh/models
2. Filter chip platform: QCS8550, Computer Vision: Object Detection, search: YOLOv5s
3. Find YOLOv5s, click to enter the detail page, view YOLOv5s-INT8-QNN236 model, check performance benchmarks
4. Log in with your developer account, click Model & Code, download the model file and code package
5. Upload to the A8550MA1 and extract

unzip YOLOv5s_Qualcomm_QCS8550_INT8.zip -d model_farm_yolov5s

6. Run the test script:

# Enter the directory
cd model_farm_yolov5s
# Execute the script
python3 code/python/run_test.py \
  --target_model ./models/QCS8550/W8A8/cutoff_yolov5s_qcs8550_w8a8.qnn236.ctx.bin \
  --imgs ./code/python/bus.jpg \
  --invoke_nums 10

AIMO Model Optimization Platform

AIMO (AI Model Optimizer) is a web-based, zero-code model conversion and optimization platform that supports converting mainstream framework models to Qualcomm QNN format, with automatic quantization and performance optimization.

Core Features

• Supports mainstream frameworks such as ONNX, PyTorch, TensorFlow, and PaddlePaddle
• Built-in Qualcomm platform-specific operator library, significantly improving model conversion success rates
• Supports automatic quantization and custom quantization, balancing accuracy and performance
• Generates optimized models and sample code that can run directly on the A8550MA1

Standard Conversion Workflow (using YOLOv8s as an example)

1. Visit the AIMO platform: https://aimo.aidlux.com
2. Upload the source model file (e.g., yolov8s.onnx)
3. Select target platform: QCS8550, target framework: QNN2.36
4. Use the Netron tool to inspect the model structure and fill in input/output node names
5. Set quantization parameters: select INT8 quantization, upload a calibration dataset (optional)
6. Submit the conversion task and download the optimized model package after completion
7. Upload to the A8550MA1 and load/run using AidLite SDK

AI Creator Visual Training Platform

AI Creator is an all-in-one AI development platform integrating data annotation, model training, and remote deployment, optimized specifically for edge applications.

Core Features

• Supports pixel-level annotation: object detection, semantic segmentation, keypoint detection, OCR, classification
• Pre-built six algorithm modules, supports multi-module serial composition (e.g., localization + segmentation + classification)
• Automatically tunes training parameters — train models without professional AI expertise
• Automatically converts trained models to QNN models adapted for the A8550MA1
• Supports one-click remote deployment to connected A8550MA1 devices