Defect Detection Using Edge AI with Amazon SageMaker Neo
Key DocumentDownload Defect Detection Using the AM57x Sitara Processor Technical Document
Complete this form to access: “Defect Detection Using the AM57x Sitara Processor Technical Document”
"*" indicates required fields
Defect Detection at The Edge with On-Device AI Using Amazon SageMaker Neo
A Texas Instruments Sitara™ AM5729 processor can be used to effectively run AI applications at the edge. The AM5729 architecture has multiple processors that can be leveraged for a variety of compute intensive tasks, such as running a MobileNetV2 neural network trained to detect defects in parts. This solution includes a camera, rotating plate, and camera modules to be inspected. Tying everything together was the an AM57x Evaluation Kit; equipt with cores to accelerate deep learning operations. The final solution is able to drive the rotating plate, bringing camera modules into view, then a variant of MobileNetV2 classifies the cameras as either pass or fail, depending on the presence of defects. MobileNetV2 was trained to classify based on specific defects using images collected from the setup. Training was performed in the cloud using AWS Sagemaker and compiled for target hardware using Sagemaker Neo.
- MobileNetV2 optimized and running on the AM57x Evaluation Kit at ~30fps
- >95% accurate at classifying defective cameras
- Video file or camera input
- Capture and Inference statistics displayed on screen
- Industrial Automation
- Video Surveillance
- Machine Learning
- Machine Vision
- Precision Agriculture
Technical Document (1)
Defect Detection Using the AM57x Sitara Processor Technical Document
Tools & Software
Ready to Get Started On Your Solution?
Contact Our Design Services Team
Our design services include full-cycle embedded product development engagements, where you own the resulting IP. We work as part of your development team and follow our proven Define, Design, Deploy development methodology.