deepinspection

Damage detection on physical assets

deepinspection.io

product info

By combining machine learning algorithms and remote inspection, we scan through terabytes of data to make sure that you have complete control over the health of your physical assets.

features

Automatic Inspection

We use unsupervised, supervised, and generative algorithms to automatically scan through billions of images.

Manual Inspection

Our digital inspectors manually remove false positives so  that you can focus on the actual damages.

Self-learning

By continuously annotating images and retraining the algorithms, the solution becomes almost automatic with only a tiny amount of expert supervision.

Integration&Reporting

We integrate our software into your systems to run algorithms in real–time and report damages directly to your asset management system.

algorithm

In the current version of the product we combine several open-source models to extract information. Right now, we use CLIP, FairFace, MTCNN, and Yolo.

NextDiversity works for visual data and is trained on 1TB of visual content. It can be used to work with any type of visual content with almost zero (0,0043%) of error rate.

opportunities

This product works best for any use–case where damages are rare but critical.
  • keep the inventory of your assets with no hand work
  • prevent asset breakdown for uninterrupted workflow
  • save billions by reducing the system downtime
  • perfect your maintenance schedule
  • improve your logistics and supply planning processes

customer cases

See the example of the actual usage of the same algorithm.

Detecting Damage on Catenary Wires

A broken catenary wire can stop the train traffic for hours and cost millions. With Infranord and their measurement train, we use DeepInspection to scan 20 000 km of railway in Sweden. In 2021, we analyzed over 500 M images and found more than 150 damages.

Identifying Cracks in Sleepers

Small cracks in concrete sleepers can soon evolve and cause dislocation of sleepers and the rail. In extreme cases, it can lead to derailments. To avoid this outcome, we use DeepInspection to detect and classify cracks early in their development to make sure cracked sleepers are changed.