Error Detection for Tortillas Using Deep Learning

Challenge

The inhomogeneous topology of each tortilla surface, characterized by uneven textures and patterns, can confuse conventional vision systems. Variations in shade and shaped are caused by differences in baking levels and ingredient composition, which further complicate inspection. On top of these issues, cracks and wrinkles are often subtle, blending in with normal surface textures and making them difficult to detect using standard algorithms.

Solution Implementation

senswork created a solution by using Deep Learning optical inspection with ViDi Detect. With a training dataset of approximately 100 tortilla images, the system was taught to recognize both normal product variations and specific defects. The technology focused on two primary tasks: detecting cracks on the tortilla surface and identifying wrinkles that indicate folding or processing inconsistencies. By learning directly from labeled examples, the model developed the ability to tolerate natural differences in shade and shape while maintaining accurate detection performance. Once deployed, the inspection system operated in real time, delivering immediate pass/fail results for each tortilla without slowing down production.

Benefits

• Durable inspection that maintains consistent performance over time.
• Tolerance for natural variations in shade and shape, reducing false rejections.
• Reliable deliverables that ensure customers receive consistently high-quality products.

Outcome

The implementation of ViDi Detect for tortilla inspection improved the overall quality control process. By automating the inspection process and minimizing the reliance on manual checks, production efficiency improved, resulting in consistent high-quality deliveries to customers.

For more examples of machine vision in the food industry, click here.