deepsense Quality – where an “out-of-the-box” solution is not enough

Food and Beverages

Example use cases

Correct label detection

Correct shape/orientation verification

Detection of contamination presence

Topping/sauce cover defects detection

Surface defects (e.g., cracks, holes) in the product

Product spills detection

Anomaly detection

Moisture presence detection

Correct product arrangement detection

Selected case study

Topping/sauce cover defects detection | Detection of contamination

ChallengeClient asked us to tackle 2 of if their problems on the food production line – sauce smears on the inner packaging of the product as well as detection of correct positioning of the topping of the product.

ApproachOur solution is based on the state-of-the-art detection neural networks. The system has been trained so it is not vulnerable to varying lighting conditions and can recognize the context of the scene, e.g. trigger the alarm when the topping is missing in the central part of the product and not on the edges (subject to the client requirements).

ResultsThe model, was able to identify 99% of faulty products with topping defects and raised 99% correct alarms for sauce smears. The model significantly reduced the need for manual quality control, hence lowered cost.

Automotive

Example use cases

Correct label detection

Correct shape/orientation verification

Defect area marking

Surface air bubbles detection

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Rust detection

Selected case study

Surface defects detection | Suface air bubbles detection

ChallengeDetecting a broad class of tire imperfections, including discolorations, cracks and bubbles.

ApproachSpecifically designed neural network architectures working on high-resolution 2d and 3d imagery.

ResultsPrecise assessment of defect type, severity and location – even in case barely detectable by a human expert.

Industrial Goods

Industrial goods

Example use cases

Correct label detection

Correct shape/orientation verification

Defect segmentation

Surface air bubbles detection

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Surface anomaly detection

Selected case study

Surface anomaly detection

ChallengeDetecting various kinds of flaws and anomalies in carbon fiber sheets.

ApproachCombination of detection and segmentation neural networks.

ResultsPixel-tight accuracy in finding the location, type and surface area of defects.

Packaging

Example use cases

Correct label detection

Correct shape/orientation verification

Detection of contamination presence

Defect segmentation

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Anomaly detection

Selected case study

Surface defect detection | Defect segmentation

ChallengeDetecting flaws, damages, and scratches on packages to improve packaging quality control.

ApproachSolution based on U-net convolutional neural network for image segmentation.

ResultsClose to 90% accuracy in detecting damaged surface with pixel-level precision.

Other

OtherWe’re interested to learn about your problem. We have experience working in a number of industries solving problems including…

Surface irregularities detection (e.g., hole, crack, dent)

Contamination detection

Defect classification by type

Packaging damage detection

Defect segmentation

Product orientation verification

Product completeness checking

Label detection

Defect severity assessment

Food and Beverages

Example use cases

Correct label detection

Correct shape/orientation verification

Detection of contamination presence

Topping/sauce cover defects detection

Surface defects (e.g., cracks, holes) in the product

Product spills detection

Anomaly detection

Moisture presence detection

Correct product arrangement detection

Selected case study

Topping/sauce cover defects detection | Detection of contamination

ChallengeClient asked us to tackle 2 of if their problems on the food production line – sauce smears on the inner packaging of the product as well as detection of correct positioning of the topping of the product.

ApproachOur solution is based on the state-of-the-art detection neural networks. The system has been trained so it is not vulnerable to varying lighting conditions and can recognize the context of the scene, e.g. trigger the alarm when the topping is missing in the central part of the product and not on the edges (subject to the client requirements ).

ResultsThe model, was able to identify 99% of faulty products with topping defects and raised 99% correct alarms for sauce smears. The model significantly reduced the need for manual quality control, hence lowered cost.

Automotive

Example use cases

Correct label detection

Correct shape/orientation verification

Defect area marking

Surface air bubbles detection

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Rust detection

Selected case study

Surface defects detection | Suface air bubbles detection

ChallengeDetecting a broad class of tire imperfections, including discolorations, cracks and bubbles.

ApproachSpecifically designed neural network architectures working on high-resolution 2d and 3d imagery.

ResultsPrecise assessment of defect type, severity and location – even in case barely detectable by a human expert.

Industrial Goods

Industrial goods

Example use cases

Correct label detection

Correct shape/orientation verification

Defect segmentation

Surface air bubbles detection

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Surface anomaly detection

Selected case study

Surface anomaly detection

ChallengeDetecting various kinds of flaws and anomalies in carbon fiber sheets.

ApproachCombination of detection and segmentation neural networks.

ResultsPixel-tight accuracy in finding the location, type and surface area of defects.

Packaging

Example use cases

Correct label detection

Correct shape/orientation verification

Detection of contamination presence

Defect segmentation

Sealing verification

Surface defects (e.g., cracks, holes) in the product

Process residue detection

Moisture presence detection

Anomaly detection

Selected case study

Surface defect detection | Defect segmentation

ChallengeDetecting flaws, damages, and scratches on packages to improve packaging quality control.

ApproachSolution based on U-net convolutional neural network for image segmentation.

ResultsClose to 90% accuracy in detecting damaged surface with pixel-level precision.

Other

OtherWe’re interested to learn about your problem. We have experience working in a number of industries solving problems including…

Surface irregularities detection (e.g., hole, crack, dent)

Contamination detection

Defect classification by type

Packaging damage detection

Defect segmentation

Product orientation verification

Product completeness checking

Label detection

Defect severity assessment

Get in touch

If you wish to learn more about out offer, products or possibilities to cooperate, just leave your email address. Our team will reach out with all the information you need.