Unsupervised Learning from Narrated Instruction Videos
This addresses the challenge of extracting structured task knowledge from unstructured video data for applications in robotics or education, but it is incremental as it builds on existing multimodal learning approaches.
The paper tackles the problem of automatically learning main steps from narrated instruction videos, such as changing a tire, by developing an unsupervised method that clusters video and text modalities with joint constraints, and demonstrates its ability to discover and locate steps on a new dataset of 800,000 frames across five tasks.
We address the problem of automatically learning the main steps to complete a certain task, such as changing a car tire, from a set of narrated instruction videos. The contributions of this paper are three-fold. First, we develop a new unsupervised learning approach that takes advantage of the complementary nature of the input video and the associated narration. The method solves two clustering problems, one in text and one in video, applied one after each other and linked by joint constraints to obtain a single coherent sequence of steps in both modalities. Second, we collect and annotate a new challenging dataset of real-world instruction videos from the Internet. The dataset contains about 800,000 frames for five different tasks that include complex interactions between people and objects, and are captured in a variety of indoor and outdoor settings. Third, we experimentally demonstrate that the proposed method can automatically discover, in an unsupervised manner, the main steps to achieve the task and locate the steps in the input videos.