PowerAI Vision & Other Solutions on an AI Box
summary
A low‑cost AI box is built from a 4‑core IBM Power9 CPU, 32 GB ECC RAM and an NVIDIA GeForce RTX 2070, costing about ~$3k build cost. Using PowerAI Vision it trains an MNIST model of 60,000 images, 15–20 minutes for a deployable image‑classification API; swapping to a different motherboard cuts training time by roughly 65%. Demos include automated attendance, PPE detection, retinal/CT screening; issues reported are occasional high‑confidence misclassifications and UI/coding friction.
AI-written summary · about
chapters (9)
- 0:00Introduction and team
- 1:25PowerAI Vision and AI Box
- 3:37MNIST handwriting demo
- 5:05PowerAI Vision tool walkthrough
- 9:22Performance comparison and applications
- 12:40Healthcare and industrial safety uses
- 14:06Attendance with facial recognition
- 15:33Future plans and collaborations
- 16:55Q&A session
related talks
19mCOVID Detection - Deep Learning Solution on power 9 AI Box
38mReal-world Machine Learning with TensorFlow and Cloud AI
13mAccelerating AI algorithms
01h 02mAI on the Edge
25mMaking AI Useful
30mCloud? No Thanks! I’m Gonna Run GenAI on My AI PC
02mAI Tech's Smart City solution
36mAI-Powered Real-Time Quality Control