MementoA key capability of an intelligent system is deciding when events from past experience must be remembered and when they can be forgotten. Towards this goal, we develop a predictive model of human visual event memory and how those memories decay over time. We introduce Memento10k, a new, dynamic video memorability dataset containing human annotations at different viewing delays. Based on our findings we propose a new mathematical formulation of memorability decay, resulting in a model that is able to produce the first quantitative estimation of how a video decays in memory over time. In contrast with previous work, our model can predict the probability that a video will be remembered at an arbitrary delay. Importantly, our approach is multimodal, combining visual and semantic information to fully represent the meaning of events. Our experiments on two video memorability benchmarks, including Memento10k, show that our model significantly improves upon the best prior approach (by 12% on average).
Memento10k Dataset
We present a new multi-temporal, multimodal memory dataset of 10,000 video clips, Memento10k. With 900,000 human memory annotations at different delay intervals and 50,000 captions describing the events, it is the largest repository of dynamic visual memory data.
Paper
Demo
Examine the memorability of scenes within a video using our online demo. Simply select a video from our list of examples or upload your own.
