Improving CCTV surveillance images

Examples model 2 run on large dataset (SCFace)

Deep learning models


expert workflow

Deep learning models

We gebruiken geavanceerde deep learning-modellen om de best mogelijke resultaten voor uw oplossing te bereiken. Een afbeelding kan worden verbeterd of hersteld, afhankelijk van het model dat u kiest.

Please read our disclaimer as this technique will add pixels that did not exist in the source image.


U kunt al onze deep learning-modellen zelf uitvoeren. U kunt het beste resultaat kiezen uit de meerdere uitgangen. 

After each image you can check if you desire manual post processing. (see example above)

Expert workflow

Om de beste resultaten te verkrijgen, hebben sommige afbeeldingen de deskundige workflow nodig. Dat betekent dat onze expert uw afbeeldingen voor elke stap in het proces optimaliseert. 

Manual optimization with Photoshop and other tools in pre- and post processing to create a better end result.

Frequently asked questions

About the Deep Learning models

Hoe evalueren we de nauwkeurigheid van de reconstructie?

This remains difficult with real world images as the verifying source image is obviously not present. Therefore we output result from several models. Its then up to the user to choose which fits best.

Wat zijn de toekomstige ontwikkelingen om de nauwkeurigheid van de reconstructie te vergroten?

We are constantly reviewing new studies in computer vision to stay on top of the latest technology in the field of image restoration and optimization. We are working on more than one approach to present a certain accuracy with each output result. 

One  alternative could be to use a downstream vision task e.g. face recognition, to optimize the super-resolution model. But that requires a clients database access where we will face privacy restrictions.

About the manual editing of images

Hoe behoud je de nauwkeurigheid van de reconstructie terwijl je de afbeelding handmatig bewerkt?

As stated above. Our users have to determine what output fits there need best. Based on that result we edit the image manually. In general no facial structures are added manually. Mostly superficial modifications like skin structure, hair, sharpness, brightness, colorization, etcetera.

How can we be involved in the editing process?

In an expert workflow there is the option to review stages of results.

Verbeteren jullie ook videobestanden?

We currently do not offer to restore video files. However, we can assist in frame extraction for the best results for restoring your images.

About the legal side of image restoration for law enforcement

Is een herstelde afbeelding wettelijk bewijs?

Write the answer to the question here. This way, your visitors can easily pick the questions that matter to them, without being distracted by loads of text from the FAQ.

Welke privacywetten zijn van toepassing op herstelde afbeeldingen?

Countries have different privacy laws, therefore we cannot answer this question. To determine the privacy laws applicable in your country its important to know that a restored image contains new pixels (data) dat have been added by our software that are not present in the source image.

Is er een bewezen nauwkeurigheid beschikbaar?

At this point not. However we aim to offer only "state of the art" technology in the field of image restoration with deep learning models.

Drop us a line below