The company has a rival in Polarr, which last year raised $11.5 million for its offline, on-device computational photography that’s used by companies including Qualcomm, Oppo, and Hober. It supports the deployment of model versions to test devices while training and tracks the configurations, and it protects models from attackers while improving those running on-device by analyzing platform, device, and processor performance.Īmong Fritz’s customers are Momento, One Bite, Video Star, PlantVillage, MDacne, Superimpose X, and Instasaber, who’ve used its workflows to develop models that change hair color in real time, replace photo backgrounds, identify food like pizza, detect pets, and create stickers for messaging apps. Plus, Fritz’s prebuilt models don’t require an internet connection, and they run atop live video with “a fast frame rate.” All of them optionally perform inference on-device and come in several sizes, from small models tailored for size and bandwidth to fast models optimized for processing speed.įritz enables customers to generate synthetic data or collect data for annotation and to benchmark on-device model performance before deployment.
Apple and Google offer mobile machine learning solutions in Core ML and ML Kit, respectively, but they’re platform-specific.
0 kommentar(er)
