Object Detection Model Centroid-Based Training
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection models require labor-intensive and time-consuming manual marking of training target object regions, leading to high labor costs and low accuracy in prediction results.
Innovation Solution
A method where a training target region in a training sample image is not accurately marked, using a first region and a second region determined by a sample centroid, with an object detection model calculating relation degrees for pixels within these regions to adjust the model parameters until convergence, allowing for efficient and accurate target object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual marking of training target object regions is performed accurately, then detection accuracy is improved, but labor cost and time consumption increase significantly
Solution Approach 1:
The system enables self-service by allowing the object detection model to automatically learn from images with only centroid markings. The model performs self-training by calculating relation degrees between pixels and centroids, and adjusting its own parameters through loss calculation and optimization, eliminating the need for manual region marking while maintaining detection accuracy
Solution Approach 2:
The invention changes the training parameter from requiring complete region boundaries to only requiring centroid coordinates. This parameter simplification reduces marking complexity while the model compensates through learning spatial relationships between pixels and centroids, achieving accurate detection without time-consuming boundary marking
2Measurement precision
If manual marking of training target object regions is performed accurately, then detection accuracy is improved, but labor cost increases
Solution Approach 1:
The model performs self-service by automatically learning from simplified centroid-based markings. It calculates relation degrees and performs self-adjustment through loss optimization, eliminating the need for expensive manual region annotation while maintaining high detection accuracy
Solution Approach 2:
The invention replaces expensive, time-consuming manual region marking with cheap, simple centroid point markings. These minimal markings serve as sufficient training data, dramatically reducing labor costs while the model learns effective detection through automated relation degree calculations
3Measurement precision
If manual marking of training target object regions is performed, then model training is enabled, but marking results are easily affected by marking person variability
Solution Approach 1:
The system eliminates human variability by enabling the model to learn from objective centroid coordinates without manual region interpretation. The automated relation degree calculation and loss optimization process ensures consistent training results regardless of who performs the initial centroid marking
Solution Approach 2:
The invention changes the marking parameter from subjective region boundaries to objective centroid coordinates. This parameter standardization eliminates marking person variability while the model compensates for the reduced information through automated spatial relationship learning and relation degree calculations
4Measurement precision
If accurate region marking is required for training, then detection precision is improved, but productivity decreases
Solution Approach 1:
The model achieves self-service by automatically learning from centroid-only markings through automated relation degree calculations and loss optimization. This eliminates the need for time-consuming accurate region marking while maintaining detection precision, dramatically improving training productivity
Solution Approach 2:
The invention changes the training parameter from detailed region boundaries to simple centroid coordinates, reducing marking time while the model compensates through automated spatial relationship learning. This parameter simplification maintains detection precision while significantly improving training efficiency and productivity
Data Source
AI summary
This application relates to a target object detection method and apparatus, a non-transitory computer-readable storage medium, and a computer device. The method includes: obtaining a to-be-detected image; inputting the to-be-detected image into a target object detection model; generating, by the target object detection model, a prediction diagram corresponding to the to-be-detected image, the prediction diagram describing a relation degree to which pixels of the to-be-detected image belong to a target detection object; and performing region segmentation on the prediction diagram to obtain a target detection object region. In addition, a method and an apparatus for training an object detection model into the target object detection model, a non-transitory computer-readable storage medium, and a computer device are also provided.