Video Thumbnail Selection Using Item-Image Similarity Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing e-commerce systems struggle with automatically generated video thumbnails that fail to accurately represent items, leading to inefficiencies and increased manual effort from sellers.
Innovation Solution
A computer-implemented method and system that uses a machine-learning model to determine weighted visual similarity between item images and video frames, selecting a thumbnail based on similarity scores and frame positions within the video.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic thumbnail generation is used based on predetermined video frames, then productivity is improved, but manufacturing precision deteriorates because the extracted frames do not accurately represent the item
Solution Approach 1:
The patent replaces traditional mechanical/frame-based thumbnail selection methods with a machine learning model that performs visual similarity comparison. Instead of relying on predetermined video frames or simple quality thresholds, the system uses an automated visual recognition system to identify frames that accurately represent the item, thereby maintaining high productivity while improving thumbnail accuracy.
Solution Approach 2:
The patent changes the selection criterion from temporal parameters (predetermined frame positions) or simple quality metrics to a similarity-based parameter that measures visual correspondence between video frames and item images. This parameter transformation enables the system to select frames based on their representational accuracy rather than their position or raw quality, resolving the contradiction between automation and precision.
2Manufacturing precision
If manual thumbnail generation is used, then manufacturing precision is improved, but productivity deteriorates due to time-consuming tasks for sellers
Solution Approach 1:
The patent implements a self-service system where the thumbnail generation process is automatically performed by the platform using machine learning models. Sellers no longer need to manually select or create thumbnails; the system autonomously analyzes video content, compares frames with item images, and generates appropriate thumbnails without human intervention, thereby maintaining accuracy while dramatically improving productivity.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the video content and the thumbnail output. This intermediary performs the complex task of visual similarity assessment and frame selection, bridging the gap between automatic processing and high-quality results that previously required manual human judgment.
3Device complexity
If simple quality threshold filtering is used, then device complexity is reduced, but manufacturing precision deteriorates because quality thresholds do not ensure item representation
Solution Approach 1:
The patent replaces simple quality threshold filtering with a machine learning-based visual similarity system. Instead of using basic computational rules, the system employs trained models that can understand visual content and determine whether a frame accurately represents the item, achieving high precision without proportionally increasing system complexity through the use of optimized deep learning architectures.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods are provided for automatically generating a thumbnail for a video on an online shopping site. The disclosed technology automatically generates a thumbnail for a video, where the thumbnail represents an item but not necessarily content of the video. A thumbnail generator receives a video that describes the item and an ordered list of item images associated with the item used in an item listing. The thumbnail generator extracts video frames from the video based on sampling rules and determines similarity scores for the sampled video frames. A similarity score indicates a degree of similarity between content of a video frame and an item image. The thumbnail generator determines weighted similarity scores based item images and occurrences of sampled video frames in the video. The disclosed technology generates a thumbnail for the video by selecting a sample video frame based on the weighted similarity scores.