Unveiling the Qualcomm RB3 Gen 2: A Deep Dive into Edge AI and Developer-Friendly Workflows
1:17 Demo and Surprises: The session features a live demo of the RB3 platform and hints at upcoming surprises.
1:47 IoT Focus with RB3 Gen 2: Two demos center around the RB3 Gen 2 development kit, showcasing its capabilities for IoT applications.
2:11 AWS Reinvent Showcase: The demos were initially presented at AWS Reinvent, highlighting the platform's use in edge AI workloads.
2:35 Integration with AWS Services: The RB3 platform can be used with various AWS services like SageMaker, AWS IoT, and RoboMaker.
3:00 End-to-End AI Workflow Demo: A practical example demonstrates training a YOLO V5 model on SageMaker, optimizing it with Qualcomm tools, and deploying it on the RB3 using foundries.io.
5:34 Foundries.io for On-Device Inference: Foundries.io provides a streamlined software stack for deploying OS and models on the device.
6:44 Multimedia SDK (IMDK): This SDK simplifies the creation of end-to-end pipelines for developers.
7:23 Foundries.io Recipes: Foundries.io offers pre-built recipes for the platform, simplifying OS deployment for developers unfamiliar with Yocto/OpenEmbedded.
8:15 Developer-Friendly Approach: Qualcomm emphasizes its commitment to creating developer-friendly platforms and tools.
13:16 Live Video Stream from RB3 Gen 2: A live video stream demonstrates the RB3 Gen 2's real-time object detection capabilities, identifying birds using a YOLO V5 model.
14:37 AWS Tutorial for Model Training: A tutorial on AWS demonstrates how to train the YOLO V5 model with a birds dataset on SageMaker.
17:49 Canonical Build for RB3 Gen 2: A new Ubuntu-based build from Canonical is now available for the RB3 Gen 2, simplifying development and deployment.
21:20 Model Deployment via Snap: The Canonical build allows developers to deploy their AI models as Snap packages.
26:21 Online Contest on DevPost: Qualcomm announces an online contest for developers building Windows on Snapdragon apps, with $20,000 in prizes.
27:14 Hackathon Highlights: Success stories from a recent hackathon highlight the platform's capabilities, showcasing innovative AI applications developed in just 24 hours.
31:36 Upcoming RB3 Gen 2 Hackathon: Another hackathon centered around the RB3 Gen 2 is planned in collaboration with Hackster.io, with 20 devices to be shipped to selected teams.
39:03 Ubuntu Documentation and Feedback: Documentation for the Ubuntu build is available, and user feedback is encouraged to improve it.
42:20 Call for App Ideas: The team seeks input from developers on the types of applications they want to see built on the RB3 platform.