Qualcomm releases IMSDK 2.0: Bridging generative AI and multimedia to accelerate the development of edge AI applications.

Beating AI Express: Qualcomm has launched Qualcomm Intelligent Multimedia SDK (IMSDK) 2.0, a unified AI and multimedia development framework for Qualcomm Dragonwing and Linux edge platforms. Designed to simplify integration of functions including cameras, audio/video, sensors, AI inference, and cloud connectivity, the SDK aims to accelerate the commercialization of products such as smart cameras, robots, drones, and industrial AI solutions. IMSDK 2.0 adds new features including Python and C++ Pipeline APIs, an application builder, generative AI inference, containerized microservices, AI programming agent capabilities, and "docs-as-code" functionality. Developers can switch AI models without rebuilding the entire multimedia pipeline via multiple inference runtimes such as QAIRT, ONNX Runtime, and TFLite. Qualcomm states that IMSDK 2.0 is built on the GStreamer architecture, paired with hardware acceleration plugins, zero-copy data transfer, and multi-stream, multi-model pipeline capabilities, supporting applications like visual AI, LLMs/VLMs, audio AI, and text-to-image. Its containerized microservices cover inference, analytics, and platform services, and support enterprise-grade connectivity with Kafka, MQTT, AWS IoT, Azure IoT, etc. Additionally, the SDK introduces AI programming agent skills that assist developers in completing pipeline configuration, model integration, compilation, debugging, and deployment via natural language. Qualcomm adds that the SDK is already being utilized by enterprises including Samsung, Amazon, and Bose for related product development.