qualcomm/ai-hub-apps

Qualcomm AI Hub Apps is a collection of sample applications and tutorials for deploying machine learning models on Qualcomm devices.

The repository addresses the challenge of running optimized machine learning inference on Qualcomm hardware by providing ready-to-use app recipes and end-to-end workflows. Each application is designed to work with models from Qualcomm AI Hub and demonstrates how to leverage device-specific acceleration through multiple inference runtimes: TensorFlow Lite, ONNX, and the Genie SDK for generative AI workloads. The apps support execution across CPU, GPU, and NPU compute units, with NPU acceleration available on Snapdragon chipsets including the X2 Elite, X Elite, and 8 Elite Gen 5 processors, supporting both floating-point and integer quantization formats.

Developers targeting Android, Windows, or Ubuntu deployment should consider this collection if they need concrete examples of on-device ML inference patterns. The repository provides open-source implementations for common tasks including image classification, object detection, semantic segmentation, and super resolution, with some applications featuring live camera integration and OpenCV support. Android apps are implemented in Java, while the Windows and Ubuntu offerings use Python. This is most valuable for teams building production applications on Snapdragon devices who want to understand best practices for model optimization and runtime selection rather than starting from scratch.

The project maintains a structured app directory organized by operating system and task type, with each application including its own build and installation instructions. Contributors are directed to a dedicated guide covering repository architecture and the process for adding new applications, indicating an intentional approach to scaling the collection.