About us

Synthetic Images

We present a service for creating huge amounts of synthetic images for object detection and segmentation tasks which you can use to train your deep neural networks on. Due to the lack of affordable, high quality data for custom object detection and segmentation tasks there are a lot of fields being unexplored. Our solution helps to generate meaningful synthetic images for training of neural networks in a short amount of time containing pixel wise annotation with a high, stable quality of annotations throughout the whole dataset. With our solution it is possible to avoid the need of collecting large amounts of hand annotated real-world data.

By combining our knowledge in the field of Deep Learning and Computer Generated Images we developed a solution which is capable of creating lots of perfectly pixel wise annotated images within 5 working days after receiving a 3D model of your object of choice. Our images are photorealistic with different camera angles, light augmentation, distractor objects, up to 2024 x 2024px export and different sources of light. You can choose to have between one and 15 annotated objects per image.

Real world datasets are prone to human error during the labeling or creation process. There is a tradeoff between labeling everything by a domain expert which is only possible to a small amount of images and outsource the labeling process which is cheaper and applicable for a bigger dataset, but might end in annotations with less quality due to errors in the labeling process.

If the future models should work in various environments as in SaaS solutions for consumers or end customers it is nearly impossible to create a sufficient dataset which represents the variety of our real world. The bigger the project and the higher the number of objects of interests the more time it takes to first plan the creation of the dataset and more resources are locked into managing quality and working on preprocessing methods due to the lack of labels.

Synthetic images offer the best trade off between quality and price for training object detection and segmentation models. Since images are created digitally on computers in a real world environment, the whole process of creating images and labels is highly scalable due to the help of cloud computing.

While creating the dataset synthetically we assure that the dataset will be unbiased due to quality control steps during the creation process and hand this report with every dataset that is sent to you. Since variance in light settings is the most crucial quality indicator for a dataset besides the objects, we assure that there will be no image with the same light setting. Making every image in the dataset unique which will cause a big impact in model performance when rolled out into production. Because our images are created by computers we not only get perfect labels on each object in the image over all images but also have pixel wise annotations and depth maps which assures that you get the best labels possible for object detection and image segmentation tasks. 

Best results can be achieved by combining a small amount of annotated real world images from the domain with a huge portion of our synthetic images. The model will have the ability to generalize due to randomness in the synthetic images and memorize to some degree the domain specific feature from the real world images.