Fabric defect detection from inspection photographs

Project Description

Our four-point inspection is manual and inconsistent between inspectors, and a missed defect that reaches a buyer costs far more than the fabric. We want a first pass on photographs from an inspection table — holes, slubs, contamination, shade variation — that flags for human confirmation rather than deciding alone. We can label a dataset from our own rejections. Shade variation is the hardest and the most valuable; please address it specifically.

Skills Required