Autonomous Defect Learning and Detection Engine for inspection system & on-board remote-controlled systems
Implementation Time:
9 months
Solution Provider: AI Singapore
Oceans.ai is a deep tech company developing an industrial autonomy platform to help asset maintenance and operations teams be able to take preventive action on their assets by the intelligent and efficient integration of autonomous robotics, IOT, computer vision, predictive AI, analytics and photogrammetry.
We are headquartered in Singapore and have operational bases in Africa, APAC, Europe and India. Our services extend across the globe from Latin America to Japan & Australia.Â
- Engineering inspection has been around for decades now. With the growth and scale and remoteness of industries, the complexity of inspections performed on these assets have increased.
- Today’s tools kits for these industries are a myriad set of disconnected svstem and processes that held asset owners and managers understand the state and condition of their assets starting with manually conducted visual inspections (GVI and CVI) of industrial assets
- While technology has evolved, the process of inspections is still significantly manual and tedious.With the advent of
drone and robotics-based technology, while some of these challenges have been addressed, a significant portion of the challenges remains
How can industries automate and improve the efficiency and accuracy in the asset visual inspection process using computer vision to achieve high precision and accuracy while saving costs and time?
The computer vision models were deployed on the DeepDIVE platform for enabling pre-filtering of visual data for human expert verification and certification.
- Trained and deployed several object detection models to detect and annotate 3 types of common defects in Oil & Gas structures ( corrosion/rust, crack and deformation)
- The models are exposed via an Application Programming Interface (API) that can be consumed by services in Oceans.AI
- Predictions from the models can be fine-tuned by human experts, and this feedback can then be used to retrain the models
Training models yield >90% improvement in precision from the baseline
Automated detect detection improved inspection reporting time by at least 20%
Time and money have been saved; less time needed to perform manual annotations; human experts spend their time verifying the annotations instead.
Improvement in data processing and classification for continuous learning and better models’ delivery.
PDF Document
This is the last content of this tab. If you do not see any resources above, it means the solution provider have not provided any resources. Feel free to contact Solution Provider for more information.
Implementation Time
9 months
Use Case Brochure
