Intelligent Defect Recognition 12-week PoC

In a 12-week PoC, Avvale proposes to investigate the creation of an automatic defect recognition tool starting from images generated by the client relating to the inspected product.

Solution is based on image analysis from the client inspection system, using ML/AI algorithms (image processing, image segmentation) exploiting AWS cloud technologies optimized for context. Amazon Rekognition custom label service will be tested to verify performances on recognizing production defects in client images.

Images must respect common photographic composition (e.g.: light, absence of shadows or reflections, lens distance, definition, sharpness, size, angle, ...).

PoC assesses the industrialization of a solution that greatly simplifies defect visualization, expanding its accessibility to a wider user base. The solution will serve as a platform for gathering user feedback on AI models, which in turn informs the ongoing refinement. It will assess the development of an objective metric for sorting items based on their quality level and the creation of a digital record ensuring the perpetual accessibility of comprehensive information about analyzed products.

Avvale proposes a perimeter Proof Of Concept to establish the effectiveness of the methodology as soon as possible and possibly follow a phase of industrialization and evolution of the system. The expected timeline for the implementation of the proposed solution is 3 months.

After this period, Avvale will decide with the client the next steps, supporting the definition of a effective roadmap.