A target detection model, a target detection method, a computing device and a storage medium

By constructing a hybrid sample set and class alignment hints, the problem of the imbalance between recall and precision in object detection models under non-homologous annotation systems is solved, and the model achieves high-quality training and detection capabilities across all categories without changing the model architecture.

CN122199902APending Publication Date: 2026-06-12CORECHENG (BEIJING) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CORECHENG (BEIJING) TECHNOLOGY CO LTD
Filing Date
2026-01-21
Publication Date
2026-06-12

AI Technical Summary

⚠Technical Problem

Existing object detection models suffer from class conflicts caused by non-homogeneous labeling systems in data collected from different datasets and projects. This leads to an increase in recall but a decrease in precision, making it difficult to achieve high-quality training and object detection capabilities across all categories without changing the model's basic architecture.

⚗Method used

By constructing a mixed sample set consisting of a first sample set and multiple non-homologous second sample sets, and combining class alignment cues, the model is trained to ensure that the positive sample supervision signal is correct and avoid the negative sample supervision signal error, thereby achieving high recall and high precision of the model across all classes.

🎯Benefits of technology

Without changing the basic model architecture, high-quality training of the model on non-same-origin category labeled data was achieved, ensuring the stability of the recall and precision of the model across all categories and the reliability of the detection performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122199902A_ABST
    Figure CN122199902A_ABST
Patent Text Reader

Abstract

The present disclosure relates to a target detection model training method, a target detection method, a computing device and a storage medium. The training method comprises: obtaining a first mixed sample set; wherein the first mixed sample set is composed of a first sample set and at least two second sample sets, the at least two second sample sets are respectively associated with different second category sets, the first sample set is associated with a first category set, and the first category set contains all categories in each second category set; and training a first target detection model based on the first category set to perform a target detection task through the first mixed sample set; wherein each sample set in the first mixed sample set is supervised in the training according to the category set associated with each sample set.
Need to check novelty before this filing date? Find Prior Art