← All posts

Build & Deploy AI Vision Models on Edge Devices with EyePop.ai and Qualcomm

Run Custom Object Detection Models Offline with Qualcomm AI Hub

EyePop.ai CPO Andy Ballester shows how to build a custom AI vision model and run it on-device with Windows + Qualcomm NPUs—ideal for high-performance, private inference without internet access.

What You’ll Learn in This Demo

EyePop.ai CPO Andy Ballester shows how to build a custom AI vision model and run it on-device with Windows + Qualcomm NPUs—ideal for high-performance, private inference without internet access.

1️⃣ Install EyePop.ai’s Windows Edge Runtime (Timestamp: 0:00)

  • Step-by-step install for Snapdragon X Elite devices
  • Connect local runtime to your EyePop.ai account
  • Run models completely offline

2️⃣ Python Edge Inference Example (Timestamp: 2:30)

  • Detect people and poses from local images
  • Run object detection locally with EyePop.ai SDK
  • Visualize bounding boxes and keypoints using Matplotlib

3️⃣ Train a Custom Model Using EyePop.ai (Timestamp: 14:44)

  • Upload and label your own dataset with auto-labeling
  • Include negative examples to improve model accuracy
  • Prepare and train in hours, not weeks

4️⃣ Optimize Model with Qualcomm AI Hub (Timestamp: 19:12)

  • Export and tailor your model for Snapdragon NPU performance
  • Target specific Qualcomm devices for optimization
  • Combine models and deploy locally for production use

5️⃣ React Edge Inference Example (Timestamp: 26:25)

  • Upload photos, videos, or stream live input
  • Visualize results instantly—no internet required
  • Detect custom objects, poses, and more from the browser

‍

Code & Developer Resources

‍

Start Building Today with Qualcomm x EyePop.ai

Set up your Snapdragon device and start building AI-powered vision apps—no cloud needed. The future of vision AI is happening at the edge. This is how to be part of it. 👉 Sign up via Qualcomm AI Hub

‍