Automated Live Video Chapter Estimation via Viewer Interaction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In conventional live video distribution systems, streamers manually generate chapter information indicating the time periods during which products are presented, which is a heavy burden, and there is a need for automated estimation of these periods.
Innovation Solution
A server device collects access information from viewer terminals during live video distribution, estimating the time periods during which products are presented based on viewer interactions with associated objects, and automatically generates chapter information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If streamers manually generate chapter information by reviewing videos to confirm product presentation times, then the accuracy of chapter information is improved, but the time and labor burden on streamers increases significantly
Solution Approach 1:
The system enables automatic generation of chapter information by having the video content itself provide the necessary data. The streamer's video is automatically analyzed by detecting product objects within it, and chapter information is generated without requiring the streamer to manually review and confirm timing, thus eliminating the time burden while maintaining accuracy through automated object detection
Solution Approach 2:
The manual mechanical process of reviewing videos and confirming product presentation times is replaced with an automated computer vision system that detects product objects in the video. This substitution uses image processing and object recognition algorithms to automatically determine when products are presented, replacing the need for human manual analysis
2Ease of manufacture
If streamers manually generate chapter information, then chapter information can be created, but the complexity and burden of the generation process increases
Solution Approach 1:
The video content serves itself by providing visual information that the system automatically processes. The streamer uploads the video and the system automatically detects products and generates chapters, making the process as simple as video upload without requiring complex manual operations or specialized knowledge
Solution Approach 2:
The system extracts only the necessary information (product objects and their timing) directly from the video content itself, separating this extraction process from the complex manual tasks of reviewing, analyzing, and documenting product presentation times. This extraction approach simplifies the overall process while maintaining comprehensive chapter information
Data Source
AI summary
A receiver (120) receives, from each of a plurality of terminals being playing a distributed live video, a piece of access information sent in response to an operation on one of a plurality of objects respectively associated with a plurality of products presented in the live video. A collector (142) collects video identification information identifying the live video, product identification information identifying a product associated with the operated object, and time information indicating a playback time of the live video at a point of time at which the operation is performed, the video identification information, the product identification information, and the time information being included in each of the received pieces of access information. An estimator (143) estimates, based on the collected video identification information, product identification information, and time information, a time period during which each of the plurality of products was presented in the live video.