Object Recognition Model for Intelligent Special Effects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mixed reality technologies are limited in their ability to utilize real-world spatial information effectively, particularly in three-dimensional spaces, and struggle to recognize objects in dynamic situations beyond simple spatial relationships.
Innovation Solution
An apparatus and method that combine object detection technology with augmented reality, using a processor to input object videos into an object recognition model, extract object images, derive weight information through a weight estimation model, and determine special effect control information based on object attributes and depth information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If object detection is performed using 2D images, then detection speed increases, but detection accuracy in 3D space decreases
Solution Approach 1:
The patent transitions from 2D image-based object detection to 3D point cloud-based detection. By utilizing depth information from sensors (such as LiDAR or depth cameras), the system creates three-dimensional representations of objects and their spatial relationships, enabling accurate detection in 3D space while maintaining real-time performance through optimized processing pipelines.
2Productivity
If augmented reality technology is used to replace real objects with virtual information, then real-time interaction is improved, but utilization of real world spatial information is limited
Solution Approach 1:
The patent introduces an intermediary processing layer that captures real-world spatial information through depth sensors and point cloud generation, then translates this information into a format suitable for augmented reality rendering. This intermediary step preserves the richness of real-world spatial data while enabling real-time virtual object placement and interaction, bridging the gap between physical and virtual environments.
3Productivity
If simple spatial relationship recognition is used, then processing speed increases, but understanding of dynamic situations decreases
Solution Approach 1:
The patent implements preliminary action by pre-processing point cloud data to extract key spatial features, object boundaries, and relationship patterns before main processing occurs. By preparing and organizing spatial information in advance through filtering, segmentation, and feature extraction, the system enables both high processing speed and comprehensive situational understanding during real-time operation.
Data Source
AI summary
An apparatus for creating intelligent special effects based on object recognition according to an example of the present disclosure includes a communication module for receiving a photographed video of an object; a memory storing a program for creating special effect information from the received video; and a processor for executing a program stored in the memory, in which the program inputs an object video acquired from a camera to an object recognition model and extracts an object image to which an attribute of each object is matched, weight information of each object is derived by inputting size information of the object image for each attribute of each object to a weight estimation model, and special effect control information mapped to each object is determined based on the weight information for each attribute of each object.