Video Object Detection Using Unified Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection technologies focus more on hand modeling rather than the object held by a person, and are not universal, failing to accurately detect diverse objects in videos.
Innovation Solution
A video processing method using semantic segmentation with a convolutional neural network to identify the image region where a person holds an object by labeling any object-in-hand as a consistent semantic class, allowing the network to detect and track objects in real-time video processing scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object detection technology focuses on hand modeling, then hand detection accuracy is improved, but detection of diverse objects in hands deteriorates
Solution Approach 1:
The patent applies universality by training the convolutional neural network with a unified 'object-in-hand' semantic class that can detect any type of object held in hands, rather than creating separate detection models for different object types. This allows a single model to handle diverse objects (cups, phones, books, etc.) while maintaining detection accuracy through consistent semantic labeling across all object categories
2Measurement precision
If specific object types are detected, then detection accuracy for those objects is improved, but universality across different object types deteriorates
Solution Approach 1:
The patent changes the semantic parameter labeling approach by using a unified 'object-in-hand' class label for all objects regardless of their specific type. This parameter change allows the network to maintain high detection accuracy through consistent labeling while achieving universality across diverse object categories, resolving the contradiction between specificity and generality
Data Source
AI summary
Provided in the embodiments of the present disclosure are a video processing method and device. The video processing method includes: determining a target image to be processed in a video; performing semantic segmentation on the target image through a convolutional neural network to obtain a first feature map, wherein the first feature map comprises a feature map corresponding to at least one semantic class; determining a target image region corresponding to the at least one semantic class in the target image according to the first feature map; wherein the at least one semantic class comprises an object-in-hand, and a training image adopted by the convolutional neural network in a training process is marked with an image region corresponding to the at least one semantic class.