top of page

Reduce human-in-the-loop 'inspection' - one step at a time

Infrastructure such as roads, train tracks, buildings, overhead or transmission lines has to be available, resilient and reliable throughout its lifetime. Keeping these performance criteria at high levels, also in light of massive demographic challenges in many countries, creates the need for automation of usually tedious and manual inspection labour.
This is where technologies like unmanned aerial vehicles (UAV) in combination with advancements in Computer Vision AI can help automate large-scale visual inspection capabilities in Europe and the U.S..

Dozens of millions of high-resolution images are already taken every year by drones and helicopters. Our contribution to this market is to enable analytics of these images with cutting-edge semantic search technology.

1. Hundreds of different visual concepts have to be detected. However, it is the long-tail objects, defects and unwanted object occurrences that are most relevant to actually automate visual inspection processes.

2. Usually, very heterogeneous data is collected as components, fauna and flora, defects or even camera equipment vary from region to region or hardware provider to hardware provider.

3. Human-in-the-loop is still widely used to tackle industry-demands of near-perfect accuracy, impeding scalability.

The Challenge

The AI in Visual Inspection of Infrastructure such as Overhead Lines is slow to build and not scalable.

1. Accelerate any particular model training task by collecting the right images with just a view clicks. Train your models from 0% - 80% with out-of-the-box search capabilities in days.

2. Finetune your training datasets with high value data points that you find with Quasara to get your models to 95+% accuracy.

3. Detect long-tail items such as defects or unwanted object occurances, use our finetuning capability to build models that find defects with 80+% accuracy.

The Solution

Semantic understanding of raw, unlabeled images unlocks visual inspection every step of the way.