Neural Network Object Instance Selection via Segmented Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional digital image editing systems face limitations in flexibility, accuracy, and efficiency when detecting and selecting objects in images, often failing to accurately distinguish between multiple instances of an object, providing imprecise object masks, and requiring significant user interaction and computing resources.
Innovation Solution
The system uses natural language-based inputs to analyze user queries and determine an optimal object attribute detection model, allowing for the accurate selection of specific object instances in images based on attributes like color, position, and size, utilizing neural networks to automatically detect and select target objects within images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object detection systems are used, then object detection functionality is provided, but the systems cannot distinguish between multiple instances of the same object class
Solution Approach 1:
The patent segments the object detection task into two distinct stages: first detecting object instances using a detection model, then classifying each instance using a classification model. This segmentation allows the system to distinguish between multiple instances of the same object class while maintaining coverage of diverse object classes.
Solution Approach 2:
The patent introduces an intermediary classification model that acts as a mediator between the detection model and the final object identification. This classification model receives detected instances as input and provides detailed class labels, enabling precise instance discrimination without limiting the range of detectable object classes.
2Productivity
If conventional neural networks are used for object detection, then automatic detection is provided, but the entire system must be retrained when faults occur
Solution Approach 1:
The patent segments the neural network system into separate detection and classification models that can be independently trained and updated. When faults occur or performance needs improvement, only the affected component needs to be retrained rather than the entire end-to-end system, significantly reducing maintenance time and computational resources.
3Measurement precision
If conventional selection tools are used, then user interaction is required for object selection, but the process is imprecise and time-consuming
Solution Approach 1:
The patent implements self-service by enabling the system to automatically select target objects based on user queries without requiring manual intervention. The classification model automatically identifies and selects the desired object instance based on learned attributes, eliminating the need for users to manually adjust selection tools while maintaining high precision.
Data Source
AI summary
The present disclosure relates to an object selection system that accurately detects and automatically selects target instances of user-requested objects (e.g., a query object instance) in a digital image. In one or more embodiments, the object selection system can analyze one or more user inputs to determine an optimal object attribute detection model from multiple specialized and generalized object attribute models. Additionally, the object selection system can utilize the selected object attribute model to detect and select one or more target instances of a query object in an image, where the image includes multiple instances of the query object.