CASE STUDY
Aerospace
Our client is a prominent global tier 1 aerospace supplier renowned for its multi-technology expertise. They sought our assistance to enhance defect inspection quality control within the domain of x-ray image analysis on their welding lines, with a specific focus on the precise detection of pores. Recognizing our proficiency in this area, they approached Mindtrace to provide a tailored solution.
Challenges
- A limited quantity of unlabelled training data (15 images)
- Seeking minute imperfections in high-resolution images
- Ambiguous definition of defective pores
Data Preprocessing
Mindtrace played a crucial role in ensuring our client's data was of the highest quality for this project. Our experts performed meticulous data preprocessing, including cleaning, normalization, and augmentation, to enhance data integrity and readiness.
Training & Testing
Mindtrace harnessed the power of its cutting-edge Brain-Sense™ platform to efficiently process the client's data and conduct comprehensive model training and testing for anomaly detection, all while working with a remarkably small dataset of just 15 testing images.
Deployment
Having undergone rigorous training and testing, our anomaly detection model was now ready for deployment, conducting anomaly detection on welding pores.
Outcome
Mindtrace has effectively demonstrated the capabilities of the Brain-Sense™ platform specifically for anomaly detection, developing and delivering a trained solution that attains an impressive 87% accuracy, despite being constrained by a limited dataset comprising a mere 15 training samples. As the model processes additional images, the accuracy will continue to increase, due to the AI brain’s capabilities to continuously learn.
Average Learning Samples
Accuracy
Day Project