System and methods for object model training

By utilizing a unique model identifier to manage the object model training process, including data preparation, training, and deployment, the system addresses the inefficiencies and accuracy challenges of existing object detection models, achieving robust and scalable object recognition.

WO2025116918A1PCT designated stage expired Publication Date: 2025-06-05CHILIWORKS LLC
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
PCT/US2023/082152
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-12-01
Publication Date
2025-06-05

AI Technical Summary

Technical Problem

Existing object detection models face challenges in complex environments with overlapping objects, diverse scales, occlusions, and varying lighting conditions, and they require significant computational resources, making them inefficient for real-time applications and edge devices.

Method used

The system and methods leverage a unique model identifier to manage and execute the object model training process, which includes data preparation, model training, and model deployment, using a pre-existing list of operations and AI/ML algorithms, optimizing computational resources and improving model management.

Benefits of technology

This approach enables efficient and robust object model training, improving accuracy and scalability, reducing computational costs, and facilitating seamless deployment across various platforms, thus addressing the limitations of current object detection technologies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2023082152_05062025_PF_FP_ABST
    Figure US2023082152_05062025_PF_FP_ABST
Patent Text Reader

Abstract

Systems and methods that leverage a unique model identifier to execute and manage the object model training process, which creates and validates a custom image recognition model and includes data preparation, model training, and model deployment. Data preparation may include retrieving labeled or unlabeled data from a database, file system, or external source for the purpose of training the model. Data preparation may also include cleaning, normalizing, and transforming the data into a suitable format. Model training may involve selecting an appropriate AI / ML algorithm, setting hyperparameters, feeding the prepared data to the model and adjusting model parameters based on evaluation results to improve performance. Model deployment may include deploying the trained model into a production environment where it can be used to recognize objects in real-time scenarios. Model deployment may also include evaluating model performance using test data to validate the model.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Computer vision machine learning model execution service

    US10453165B1