A naukri.com initiative
Googleblog
2w
326
Image Credit: Googleblog
On-device GenAI APIs as part of ML Kit help you easily build with Gemini Nano
- A set of on-device GenAI APIs, part of ML Kit, is introduced to integrate Gemini Nano in Android apps with 4 new APIs: Summarization, Proofreading, Rewriting, and Image Description.
- GenAI APIs offer easy integration, run on-device, and ensure functionality without the need for a reliable internet connection, with safety protection features added.
- The APIs are built using components like Gemini Nano as the base model, API-specific LoRA adapter models, optimized inference parameters, and an evaluation pipeline for quality assurance.
- Quality evaluation of GenAI APIs shows improved benchmark scores compared to the Gemini Nano base model, with specific scores for different tasks like summarization, proofreading, rewriting, and image description.
- Sample usage of the GenAI Summarization API is demonstrated, emphasizing features like checking feature availability, downloading necessary features, and starting summarization requests.
- Guidance is provided on how to best use GenAI APIs for different tasks like summarization, proofreading, rewriting, and image description, considering factors like the length of content and desired outcomes.
- An example of an app, Envision, utilizing the GenAI Summarization API is mentioned, showcasing how concise summaries enhance user experience, especially for those who are blind or have low vision.
- Supported devices for GenAI APIs include Android devices using optimized platforms like MediaTek Dimensity, Qualcomm Snapdragon, and Google Tensor through AICore.
- To learn more and start implementing GenAI APIs, resources like official documentation, samples on GitHub, and guidance materials are available for developers.
Read Full Article
19 Likes
For uninterrupted reading, download the app