Video Sticker Generation via Facial Emotion Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for generating stickers from videos do not effectively utilize facial emotion recognition to extract relevant video fragments, limiting the ability to create stickers that accurately represent emotional expressions.
Innovation Solution
A method and device that extract an image sequence from a person-contained video, identify emotions using pre-trained convolutional neural networks, and extract a video fragment based on emotional levels, using threshold-based searching to define the start and end points of the sticker video, ensuring the emotional content matches the intended expression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional methods are used to generate stickers from videos, then the process is simple, but the emotional expression accuracy is poor
Solution Approach 1:
The video is segmented into multiple video fragments based on emotional expression changes. The system divides the continuous video into discrete segments where each fragment contains a specific emotional expression, allowing precise extraction of emotionally relevant portions while maintaining processing efficiency.
Solution Approach 2:
Emotion recognition models are pre-trained and prepared before video processing. The system performs preliminary emotion recognition on key frames to identify emotional expression points, then uses these pre-identified points to guide the video fragment extraction process, improving both accuracy and efficiency.
2Manufacturing precision
If video fragments are extracted based on emotional levels, then the sticker quality improves, but the processing time increases
Solution Approach 1:
The system extracts only the essential emotional expression points and corresponding video fragments, rather than processing the entire video. By taking out only the relevant portions that contain meaningful emotional expressions, the system maintains high sticker quality while significantly reducing processing time.
Solution Approach 2:
The system adjusts processing parameters dynamically based on emotional expression intensity. When high emotional expressions are detected, the system increases extraction precision; when emotional expressions are subtle, it uses coarser parameters, thereby optimizing the balance between sticker quality and processing time.
Data Source
AI summary
A method and a device for generating stickers are provided. An embodiment of the method includes extracting an image sequence from a person-contained video to be processed; identifying emotions of the faces respectively displayed by each of the target images in the image sequence to obtain corresponding identification results; based on the emotional levels corresponding to the emotion labels in the identification results corresponding to each of the target images, extracting a video fragment from the person-contained video, and acting the video fragment as the stickers. The image sequence comprises target images displaying faces; the identification results comprise emotion labels and emotional levels corresponding to the emotion labels. The embodiment can extract the video fragment from the given person-contained video to act as stickers based on the facial emotion match, which can achieve the generation of stickers based on the facial emotion match.