Page 539 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

(continued)

Item Details
Model Train- IndustryGPT integrates deeply with the ViMo cloud platform, acting as an
ing and intelligent assistant for industrial AI model training. Users can now prepare 4.5: Manufacturing
Fine-Tuning data, select models, and configure parameters using natural language, even
voice commands.
A built-in industrial knowledge base ensures the system accurately under-
stands complex requirements like "scratch detection." It provides full-process
intelligent support, from recommending data augmentation to dynamically
optimizing hyperparameters.
This dramatically shortens development cycles, turning days into hours. It
boosts new user efficiency by five times and lets algorithm engineers manage
triple the projects, fostering agile AI deployment and continuous quality
improvement.

Testbeds or This technology has already formed mature application cases in international
Pilot Deploy- leading enterprises such as Carl Zeiss, Airbus, Bosch, Canon, Continental
ments Group, and Schaeffler

2 Use Case Description

2�1 Description

IndustryGPT, an industrial large model, aims to significantly boost industrial productivity.
It tackles key inefficiencies in current industrial AI quality inspection: the low efficiency of
traditional manual quality inspection, the cumbersome nature of data annotation and cleaning,
and the slow pace of model training.

IndustryGPT achieves this by revolutionizing industrial software operation and interaction.
It deploys software functions through its interaction with the SMore ViMo intelligent cloud
platform. The ViMo platform is a comprehensive industrial intelligence platform that manages
the entire process from model creation to application deployment, drawing on experience
from over 1000 specialized industry scenarios.

By leveraging its understanding of the ViMo intelligent cloud platform's technical documentation,
IndustryGPT can comprehend the machine learning model development process. When users
issue natural language commands, IndustryGPT initiates ViMo's intelligent recommendation
system. This system then automatically generates algorithm solutions and handles tasks such
as data annotation, parameter tuning, and model training.

Integrating IndustryGPT with the ViMo intelligent cloud platform directly on the production
line transforms it into the "brain" of the industrial quality inspection integrated machine. This
machine is embedded into the production line to replace traditional manual inspection,
enabling the detection and screening of faulty products.

In a specific classification detection example, the solution chosen by IndustryGPT achieved
remarkable results: it took only 5 minutes and 27 seconds with 100% accuracy and recall rates.
In contrast, manual training for the same detection took significantly longer, at 9 minutes and
31 seconds, and yielded much lower performance with only a 50% accuracy rate and 66.7%
recall rate.

503