Edge Impulse has launched new generative AI features to enhance synthetic data management on edge devices. These tools involve the integration of technologies like DALL-E, Whisper, and ElevenLabs for generating synthetic images, speech, and audio data. Enterprise users can now source data from custom large language models (LLMs), including those that are self-hosted or from other providers, with additional LLM integrations expected in the future.
These enhancements are aimed at simplifying the data creation process for various applications. For instance, DALL-E can be used for image generation, Whisper for speech recognition datasets, and ElevenLabs for producing sounds such as alarms or glass breaking. An iterative workflow allows users to refine input prompts and manage outputs effectively.
Currently available to enterprise-tier users, these features are accessible in the “data acquisition” section of the Edge Impulse platform, and the company plans to extend them to professional plan users. This integrates with Edge Impulse's existing tools, including the NVIDIA Omniverse Replicator for computer vision model training.
By streamlining the generation of synthetic data, Edge Impulse aims to support developers in creating high-quality datasets for their models, ultimately facilitating more efficient and effective model training.