DENKblatt

THE reading for your developers or product managers → DENK Docs.

Did you know that you can integrate DENKnets seamlessly into your existing systems?

Whether API, SDK or other tools - we offer you the right tools for your individual requirements. You can find the integration guide here: docs.denkweit.ai

With our detailed integration documentation, we give you everything you need: docs.denkweit.ai

▪️ Clear instructions:
   Use step-by-step explanations to quickly integrate APIs and SDKs into your software.
▪️ Code examples:
   Ready to go in minutes - copy, customise, get started.
▪️ All-round support:
   Answers to frequently asked questions and direct contact options.

So you can implement DENKnet in any system and work smarter, faster and more efficiently.

What would have happened without personal customer support?

A fairly new customer of us wanted to start the AI training before his onboarding appointment ( introduction to the DENK Vision AI Hub)... And already invested a few hours in annotating* in the DENK Vision AI Hub.

The use case: glass splinters of different colours and sizes.

As the size of the glass particles (for later sorting processes) is also important for the customer, ‘instance segmentation’ was selected as the technology. For AI training, this means that it is necessary to mark each object accurately.

The solution for time-saving AI training

Our customer started (initially) full of enthusiasm with the manual marking of each individual part, but stopped after several hours (and a few finished images) exhausted - fortunately! This allowed Jörn, our excellent customer advisor, to intervene and show him a much quicker way during onboarding: Annotating using the 1-Click-Tool.

💡 The 1-Click-Tool makes it possible to mark objects with just one mouse click! How does it work? It is based on intelligent, pixel-precise edge detection.

Within minutes, our customer can now complete the same task that previously took him a whole working day. The personal consultation came just in time for everyone involved. PS: Jörn never leaves a customer dissatisfied! 😉

Demo on the website

Just in case you are one of the ‘’unaware‘’, you can try out our one-click tool directly on our website: to the demo.

*Marking the objects that the AI should recognise and classify.

‘If you don't do it yourself ...’ I'm sure everyone who has been dissatisfied with the work of a colleague has said this at some point. 😄

This can certainly happen when training an AI.

AI training?

Open source solutions are often generalised and reach their limits (e.g. under different lighting conditions). → A trainable vision AI, on the other hand, can achieve absolute reliability & stability - regardless of the conditions. Training means: the objects to be recognised or similar are ‘marked’ on an image. The AI learns from these ‘labels’. Incidentally, 15 images are sufficient for this process 😉

‘How do you ensure that all employees are labelling the same/the right thing?’
‘How do you ensure thorough AI training?’

If several employees are responsible for training (labelling), consistent implementation and accuracy must be ensured. → Our Quality Centre in the DENK Vision AI Hub solves this problem. It provides an overview of the degree of agreement (score) between multiple user labelling OR between user labelling & AI prediction.

⚡ ‘The AI does not recognise the bark beetle in the image’. → ‘Has this pest been labelled? → No. Then catch up quickly!’

⚡ ‘The AI recognises a crack in the picture, but it hasn't been marked.’ → ‘Great that the AI has already recognised it on its own. → But it was forgotten to mark it? → Catch up quickly so that the AI stays safe.’

✔️ Careful AI training → high-performance deployment in operation.

Contact us if you have any questions!