Object Node Identification Using Cropped Target Image Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition technologies face challenges in accurately identifying and distinguishing specific object nodes, particularly in cases where the target object occupies a small proportion of the image, leading to reduced precision and efficiency in applications like hand gesture recognition.
Innovation Solution
A method involving image detection to isolate a target region, followed by image cropping to obtain a subimage, and subsequent node recognition using optimized neural networks for precise node identification, including hand gesture skeleton recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models with large number of parameters are used for fine-grained image classification, then classification accuracy is improved, but model inference time increases
Solution Approach 1:
The patent segments the image processing task into multiple stages: coarse-grained classification first, then fine-grained classification only for selected regions. This segmentation allows the system to use different model complexities for different tasks, improving overall efficiency while maintaining accuracy.
Solution Approach 2:
The patent performs preliminary coarse-grained classification on the entire image before conducting fine-grained classification. This preliminary action identifies regions of interest that require detailed analysis, preventing the computationally expensive fine-grained model from processing the entire image.
2Measurement precision
If fine-grained classification is performed on the entire image, then object identification accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent applies different processing quality to different regions of the image. Regions identified as containing target objects receive fine-grained classification (high quality), while other regions receive coarse-grained classification or no processing (lower quality), optimizing resource allocation.
Solution Approach 2:
Instead of performing fine-grained classification on the entire image (excessive action), the patent applies fine-grained classification only to specific regions of interest identified by coarse-grained classification (partial action), reducing computational overhead while maintaining necessary accuracy.
3Reliability
If multiple classification models are used for different classification tasks, then classification performance is improved, but device complexity increases
Solution Approach 1:
The patent designs the classification system with multi-functional models that can perform both coarse-grained and fine-grained classification tasks. The same model architecture is used for different classification granularities, reducing the need for separate specialized models and simplifying the overall system.
Data Source
Figure 1~2
Figure 3a~3b
Figure 4~5
AI summary
The embodiment of the present application discloses a method and device for identifying an object node in an image, a terminal and a computer readable storage medium, the method comprises: acquiring an original image to be identified, the original image comprises a target image object; performing image detection on the original image to determine a target image region in the original image, the target image region including the target image object; carrying out image cropping on the original image according to the target image region to obtain a subimage comprising the target image object; and calling a preset node identification module to perform a node identification on the subimage to obtain node identification information of the target image object.