Perception is achieved through AI networks trained on images collected worldwide and stored in the cloud for training. Our software libraries, optimized for CPUs and GPUs, enable the compression of these training sets, specifically designed to enhance ADAS (Advanced Driver Assistance Systems) models. The initial dataset can be archived, while the compressed set is used daily. Decoding occurs in real-time during training.

• Reduced Storage Costs: The amount of data needed to store images can be reduced by a factor of 5 to 10, significantly lowering storage costs in data warehouses. Large data storage on Public Cloud rapidly becomes too costly and completely unreasonable.
  
• Continuous Improvement of ADAS Models: Cost reduction helps overcome budget constraints, allowing ongoing development of ADAS models over time, ensuring more precise and higher quality models.

Images are recorded by cars traveling around the world. These cars are equipped with data loggers that store the images before they are transferred to the cloud. Our IP core hardware for FPGAs or ASICs and our software libraries can be used to compress stored images.

• Cost-Effective Storage: Compression allows for more images to be stored on data loggers, significantly reducing storage costs.
• Faster Upload Times: Compression decreases upload times to the cloud, improving overall efficiency.
• Seamless Data Increment: The training set can be incremented without needing to recompress in the cloud, streamlining the data management process.

Over time, JPEG XS RAW will become the de facto standard for automotive image transmission and can be integrated directly into image sensors using our low-power ASIC solutions. Decoding can be performed directly on ADAS platforms using our GPU libraries and/or dedicated hardware.

• Reduced Power Consumption: Reducing the data transfer rate to the central processor will decrease the power required by the camera module, eliminating overheating issues.
• Enhanced Image Quality: ECU will no longer be limited by the bandwidth of their physical interface, allowing for higher resolution, frame rates, or bit depth of images, leading to higher quality perception.