The quality of data used in AI systems is increasingly important, especially in critical applications such as medical diagnostics, autonomous vehicles and earth observation. 
Dotphoton’s Jetraw product focuses on bringing data-centric techniques to image-based AI in those fields. 

The technology is based on a traceable image data path including calibration of imaging optics and sensor parameters. 
Jetraw includes a physical model of these properties, which is used first of all to apply lossless compression to the raw data with a high compression ratio (5:1 – 10:1) and at a speed that is 10x–100x faster than classical solutions. 
This allows AI scientists to work with data of higher quality, in higher quantities, faster.

Next, this same model is used to process the raw images to generate synthetic images accurately, changing model parameters in-silico. This can be used to augment datasets according to a variety of realistic parameters, such as illumination, resolution, focus, etc.

The physical model can also be used to pre-qualify images to ensure that they are suitable for the specific algorithm.
Jetraw components are available in both software and hardware form (FPGA) and are useful anywhere along the image-processing path: from in-camera processing to the development of advanced neural networks