Edge Impulse Introduces Generative AI Tools to Create Synthetic Data at the Edge

Edge Impulse introduced new generative AI capabilities aimed at creating and managing synthetic data on edge devices, including images, speech, and audio data.

According to the company, the new Synthetic Data integration offers a way to use Edge Impulse to create data based on a large language model (LLM). It uses tools such as DALL-E to generate images, Whisper to create speech elements for keyword searching, and ElevenLabs to generate auditory events.

The company says that using the new features, enterprise customers will have the ability to incorporate custom LLM sources, including third-party data providers or self-hosted LLMs. Additional LLM toolkits are expected to be added in the coming months.

These features complement Edge Impulse’s existing integration with NVIDIA Omniverse Replicator, a platform for developing synthetic data generation pipelines for training computer vision models.

The new toolkit is currently available for enterprise users, with the company mentioning plans to expand availability to professional plan users. It’s located in the “data acquisition” section of Edge Impulse, alongside options like Dataset, Data Explorer, and Data Sources.

The company also boasts about this integration, which allows users to add and refine suggestions, and the results, such as images and audio snippets, are displayed for evaluation.

Highlights include creating image datasets using the DALL-E model, generating keyword mining datasets for speech recognition applications using the Whisper model, and generating sound events such as breaking glass or alarm sounds using the ElevenLabs Sound Effects model.

These features also connect to other LLM data providers or self-hosted LLM systems via transformation blocks, including Edge Impulse integration with GPT-4o for labeling image data.

This iterative workflow simplifies the process of generating appropriate prompts and ensures that any unremoved data is automatically added to the project, allowing for seamless management.

These new features are designed to streamline the workflow for building models using synthetic data, making it easier for developers to create high-quality datasets using generative AI.

Related

Article Topics

data | edge computing | IoT | serverless edge computing