Authors: Zekang Yang, Wang Zeng, Sheng Jin, Chen Qian, Ping Luo, Wentao Liu

Published on: February 23, 2024

Impact Score: 7.6

Arxiv code: Arxiv:2402.15351

Summary

• What is new: AutoMMLab introduces an LLM-empowered AutoML system that automates the entire model production workflow for computer vision tasks, which is a novel approach not seen in traditional AutoML systems.
• Why this is important: There is no existing AutoML system that automates the complete end-to-end model production workflow, limiting the ability for non-expert users to easily produce models for specific tasks.
• What the research proposes: The introduction of AutoMMLab, utilizing language models to automate the entire model production process for computer vision tasks, allowing for easy creation of task-specific models.
• Results: AutoMMLab proved to be versatile and effective across a range of computer vision tasks such as classification, detection, segmentation, and keypoint estimation.

Technical Details

Technological frameworks used: RU-LLaMA for pipeline scheduling and understanding user requests, HPO-LLaMA for hyperparameter optimization

Models used: Language Large Models (LLMs) for automating model development and hyperparameter optimization

Data used: A new benchmark called LAMP for evaluating the system’s performance

Potential Impact

This innovation could disrupt the computer vision market, affecting companies specializing in manual model development and providing a competitive edge to businesses adopting this automated approach.

Want to implement this idea in a business?

We have generated a startup concept here: VisionaryAI.