Digital quality control using computer visioning with deep learning

A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for quality control based on one or more samples of a product, the operations comprising:

• receiving sample data of the product comprising digital data representative of a sample of the product;
• obtaining a set of features by processing the sample data through multiple layers of a residual network;
• identifying a set of regions of the sample of the product by processing the set of features using a convolution neural network (CNN), each region of the set of regions including at least one object;
• in response to determining that a number of objects within the sample of the product satisfies a threshold,
• adjusting the sample data of the product to generate adjusted sampled data, wherein adjusting the sample data of the product comprises splitting the digital data representative of the sample of the product into multiple overlapping parts, and
• reprocessing the adjusted sample data to identify the set of regions and the at least one object in each region;
• determining a type of the at least one object in each region;
• and
• selectively issuing an alert at least partially based on the type of the at least one object in each region, the alert indicating contamination within the sample of the product.

A system, comprising:

• one or more processors;
• and
• a computer-readable storage device coupled to the one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for quality control based on one or more samples of a product, the operations comprising:
• receiving sample data of the product comprising digital data representative of a sample of the product;
• obtaining a set of features by processing the sample data through multiple layers of a residual network;
• identifying a set of regions of the sample of the product by processing the set of features using a convolution neural network (CNN), each region of the set of regions including at least one object;
• in response to determining that a number of objects within the sample of the product satisfies a threshold,
• adjusting the sample data of the product to generate adjusted sampled data, wherein adjusting the sample data of the product comprises splitting the digital data representative of the sample of the product into multiple overlapping parts, and
• reprocessing the adjusted sample data to identify the set of regions and the at least one object in each region;
• determining a type of the at least one object in each region;
• and
• selectively issuing an alert at least partially based on the type of the at least one object in each region, the alert indicating contamination within the sample of the product.