Automated quality control
Label inspection portal
This project is a web-based image inspection system. Users upload images of the actual produced labels. The system automatically verifies it against a template using a pretrained ResNet18 model.
Templates
Upload a reference template with a friendly name and default tolerance, then select it when you upload an image for inspection.
Processing
Upload an image of the label to be inspected and it will be sent to ML service that will calculate a similarity percent having the template as a basis.
Outcome rule
The tolerance defines the deviation from the 100 % match that is acceptable and the product still can be marked as valid. Template base tolerance seeds the image form, but users can override it while uploading new image of a label.