Dual Model Image Classification Merging for Video Censorship
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image classification systems face challenges in accuracy and speed when dealing with video censorship, as they require long update cycles and slow model updates to retrain models upon detecting variant samples.
Innovation Solution
The method involves using two machine learning models: a first model for initial classification and a second model for vector search-based classification. The results from both models are merged to determine a target classification result, improving accuracy and speed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single machine learning model is used for image classification, then the system is simple and easy to maintain, but the classification accuracy is insufficient and update cycles are long
Solution Approach 1:
The patent combines two different machine learning models (a first model for initial classification and a second model for vector search-based classification) to process images. The first model provides baseline classification while the second model refines results using vector search, merging their outputs to achieve higher accuracy without requiring complete model retraining
2Reliability
If the model is retrained and redeployed upon detecting variant samples, then the classification accuracy improves, but the update cycle becomes long and model updates slow
Solution Approach 1:
The system performs preliminary classification using a first machine learning model that can quickly identify potential issues. When variant samples are detected, the second model activates to perform vector search and provide refined classification results, enabling faster response without requiring complete model retraining and redeployment
Solution Approach 2:
The patent introduces an intermediary classification result determination process that synthesizes outputs from both models. This intermediary layer allows the system to incorporate new information from vector search results without requiring full model retraining, thus improving reliability while reducing update time
Data Source
AI summary
The present disclosure provides methods and apparatuses for image processing and image classification. In one embodiment, the method for image processing comprises: receiving an image; obtaining a first classification result for the image based on a classification model; processing the image for classification based on a preset process, and providing a processing result into a re-ranking model to obtain a second classification result for the image; and determining a target classification result for the image, based on the first classification result and the second classification result.