�

TimesAI deep learning development platform independently developed by Dihuge
Provide development tools for the entire inspection process from image acquisition to model inspection to production on-line
The development mode that supports zero-code programming can quickly build machine vision application systems

AI large scale model and small model cooperation

Using industrial AI large scale models to generate defect samples, intelligent labelling, and enrichment of defect libraries.
Machine Vision Algorithm Library

Rich machine vision algorithm toolkits, flexible invocation methods, and good performance.
Deep Learning Training Platform

Using a small number of samples, an industrial Al depth learning visual detection model suitable for specific application scenarioses can be trained.
Multi-scale Product Defect Library

Deeply engaged in the field of Al vision for many years, we have a large amount of industrial product defect library.
Industrial Equipment Communication Library

Support industrial equipment such as industrial cameras, 3D scanners, PLCs, and protocols such as Modbus, TCP, and serial communication.

Core Algorithm Functions

Machine Vision+Deep Learning

Dimensional Inspection

Measure product dimensions through positioning and quantitatively analyze product dimensions.
Object Detection

Target positioning and classification of products, commonly used for small target detection, multi-target detection, etc.
Segmentation Detection

Pixel-level inspection of products, often used for product surface scar inspection, etc.
Classification detection

Classification and inspection of products to determine whether they are good crystals or defects, often used for category judgments, etc.

3D Vision+Positioning Guidance

3D detection

By analyzing and transforming the point cloud data, the 3D information of the object is extracted and quantitatively analyzed.
Positioning and assembly of equipment workpiece

Using the correlation fusion of 2D/3D information to achieve module assembly, inspection material positioning, placement, and appearance inspection, etc.

Platform Advantages

High precision

3C inspection Highest precision = 0.0005mm
Lithium-ion inspection Highest precision = 0.05mm
Functional defects Leakage rate = 0%
Non-functional defects Leakage rate<0.1%
All defects Minimum overkill rate <1%
High complexity

Support multi-model common line
3C inspection support one-key change type
Li-ion battery soft pack cell inspection support within 30min change type

High performance

Support 3C inspection CT < 0.3 s/pcs
High reusability

Reuse the previous generation of deep learning models for defect detection, supporting detection under the condition of 0 defective samples for new products.
Fast model training

Use AI large scale models to batch generate or label defective images, enrich the detection of small models, quickly complete the model iteration, shorten the project cycle

Application Demonstration

01

Hardware debugging

Integrated, unified and visualized, camera, hardware interface tools, multi-camera scheduling debugging tools to make the debugging process more convenient
02

Master development

In response to the diverse characteristics of industrial scenarioses, TimesAl supports flexible and freely configurable display units, flow meter indicators, results query, export and upload logic
03

Algorithm development

Vision algorithm module with balance of generality and ease of use, visual debugging process, pre-trained defect library for different materials, developers can achieve 0 code algorithm to go online quickly
04

Algorithm tuning

Fast model iteration based on continuous learning, interpretable/self-learning tools for industrial scenarioses, dramatically reducing model optimization time
05

Intelligent production management

Provide a variety of production data statistics templates, such as workpiece capacity statistics template, yield analysis template, etc., to achieve real-time supervision of production data

Version

Service

Based on the TimesAl deep learning development platform, Dihuge provides semi customized algorithms for visual inspection projects. Customers can enjoy the following 2 services:

Dihuge is responsible for the detection rate of the platform and secondary development algorithm;

Dihuge provides after-sales consulting services for optical imaging effects and overall debugging, including on-site technical support, remote technical support, etc;

�

01

02

03

04

05

�