What is AI Creator?
A visual model training platform that integrates labeling, training, and deployment.
AI Creator focused on realizing closed-loop management of data management, pixel-level labeling, training, and remote deployment. It helps users efficiently and quickly complete the training and updating of AI models.
AI Creator is equipped with various training algorithms, enabling model training with minimal sample data, suitable for various AI scenarios. Additionally, it integrates Abnormal training functionality, allowing the model to quickly add samples and improve accuracy.
The pixel-level labeling feature of AI Creator makes data annotation more precise, significantly enhancing model quality. All operations are visualized through the interface, greatly simplifying the training process.
Features
Built-in Six Algorithm Modules
AI Creator supports training for six types of algorithms: localization, segmentation, Abnormal, object detection, OCR, and classification.
Supports Multiple Module Combinations
AI Creator supports serial combinations of multiple algorithm modules like: localization + segmentation, localization + Abnormal, localization + segmentation + classification, and more.
Preset Training Parameters
AI Creator presets training parameters for each algorithm module and automatically selects the appropriate parameters based on sample size.
Visualized Training Process
Users can view and terminate the model training progress in real-time and compare test results with labeled results for quicker model evaluation.
Built-in Model Optimization
AI Creator includes AI model optimization tools (AIMO) that support one-click optimization of x86 platform models to Qualcomm platform (qnn, snpe) formats.
One-Click Smart Camera App Generation
AI Creator can generate smart camera applications from trained models with one click, running them on smart camera devices without coding to view model effects.
One-Click Deployment
AI Creator can be integrated with smart cameras and AidLux with one click, allowing models to be actively pushed to devices upon integration.
AI Creator Workflow
Data Preparation Stage
AI Creator supports flexible data import methods. Users can import collected samples and use AI Creator for online labeling. It also supports importing labeled data from mainstream annotation tools (PixelAnnotationTool, LabelMe).
Training Stage
In the training stage, after completing data annotation, the model training begins. Training logs can be viewed in real-time, and an automatic model evaluation occurs once training is complete, allowing you to check model performance in testing.
Model Optimization, Application Generation, and Deployment
A standout feature of AI Creator is its ability to optimize trained models to the Qualcomm platform with one click, completing deployment and application generation.