The MasterpieceAI system, independently developed by Bozhon Precision Engineering, is officially launched. The MasterpieceAI system is an industrial vision deep learning platform software that integrates data analysis, annotation, parameter tuning, training, and testing. Through a flexible and simple GUI interface, users can quickly create deep learning application systems to meet the needs of visual inspection, classification, positioning, and other applications.
Artificial Intelligence, Defective Image, Defective Sound, 3D Multidimensional Big Data Analysis for Quality Inspection. This system includes four major project modules: Project Management, Sample Annotation, Build & Train & Test Model, and Model Evaluation.
Additionally, it includes six sub-functional modules: Image Segmentation Algorithm, Image Classification Algorithm, Object Detection Algorithm, Single Sample Detection, OCR Character Recognition, and Target Measurement Identification. Through training and modeling of images and characters, it accurately segments targets.
High precision, fast speed, strong versatility.
The MasterpieceAI system utilizes the most cutting-edge deep learning algorithms to achieve higher detection accuracy and precision; it processes data quickly and has stronger versatility. While developing and making breakthroughs in deep learning algorithms, it also closely follows the research directions of non-deep learning machine learning and other AI algorithms. With a small amount of training samples, it can achieve very high accuracy, solving the difficulty of factories in obtaining sample data.
At the same time, the MasterpieceAI system has five major advantages: strong compatibility, customizability, modular functionality, high performance, and a comprehensive set of features.
MasterpieceAI Product Advantages
Strong Compatibility: The software can be flexibly deployed on various computing chips, such as CPU, VPU, GPU, FPGA, without being constrained by hardware devices. It can also achieve multi-hardware, heterogeneous hardware platform accelerated processing, improving the utilization of hardware resources.
Customizability: In terms of training models, the software can offer more choices, including algorithms, model layers, custom functional modules, formula screening, etc. It supports customizing specific algorithms for projects, providing a more flexible and vertical solution to users' fundamental needs.
Modular Functionality: In project processing, users can customize the logical order of algorithm modules. To apply technology conveniently and quickly to the production line, users can flexibly adjust and combine multiple solutions.
High Performance: Unlike ordinary AI vision inspection software, MasterpieceAI is equipped with the latest deep learning framework. After targeted optimization in algorithms, it not only ensures high precision and accuracy of defect detection but also optimizes the loss function and related hardware computing capabilities, coupled with hardware parallel processing to enhance the comprehensive detection speed of the software.
Comprehensive Features: The software has a simple style and comprehensive functions, with built-in rich data processing capabilities and a large number of adjustable parameters, allowing users to choose independently based on the project. In the segmentation algorithm, targeted reinforcement training is integrated, and in the classification algorithm, a heatmap that can assist in debugging is incorporated.
Traditional Vision Algorithm Module: While opening up deep learning as the core algorithm function, the software also integrates multiple traditional machine vision algorithms, adding more image processing tools. In positioning and measurement, it provides more multidimensional image processing methods.
Rich Sample Post-processing: It offers a wealth of data processing tools, allowing users to flexibly use these tools for prediction data calculation and filtering, review model results, and evaluate model performance through multiple indicators, observing the trend of model optimization in a more conspicuous way.
Cross-Operating System and Hardware Platform: The software can flexibly run on various computing chips without being constrained by hardware devices. It can achieve heterogeneous hardware parallel processing. Application Scenarios: Not limited to NVIDIA GPUs, it also supports Intel VPU, CPU, FPGA/integrated graphics, and supports Windows, Linux, Mac OS systems.
Visualization to Open the Black Box: Visualization can bridge the gap between abstract data and intuitive representation well. When implementing visualization for scene classification models, an optimization process is proposed based on the visualization model to improve the model's classification capability by reducing internal dataset biases. It provides a focus display for neural network training, facilitating engineers in debugging and enhancing the neural network detection effect.
Small Sample Data Training: When individual defect types lack sufficient image data, algorithms such as data augmentation are used to simulate rich data volumes, compensating for the shortcomings of insufficient sample quantities. Application Scenarios: Used for products with a shortage of individual defect type images, incomplete recognition of defect types, etc.
Transfer Learning: Transfer learning can transfer models trained on large datasets to multiple domains with only small datasets, reducing training costs and saving training time. Application Scenarios: Can be used in product upgrades or changes in product models, or in parallel use of different products with similar materials, etc.
The MasterpieceAI system is applied and deployed on the production line, allowing users to quickly establish the logical order of custom algorithm modules in the actual production line. They can freely combine and adjust multiple solutions, swiftly completing the deployment work online.
System Configuration Requirements
Minimum Specifications: Compute Capability 5.2 or higher, at least 4G of remaining storage space;
Recommended Specifications: Compute Capability 7.5 or higher, at least 8G of remaining storage space.
Wechat scan
Official video Number
Back
No.666, Huxin Road, Wujiang District, Suzhou, Jiangsu, China 
