Adversarial Unsupervised Video Summarization Augmented With Dictionary Loss

Abstract

Automated unsupervised video summarization by key-frame extraction consists in identifying representative video frames, best abridging a complete input sequence, and temporally ordering them to form a video summary, without relying on manually constructed ground-truth key-frame sets. Stateof-the-art unsupervised deep neural approaches consider the desired summary to be a subset of the original sequence, composed of video frames that are sufficient to visually reconstruct the entire input. They typically employ a pre-trained CNN for extracting a vector representation per RGB video frame and a baseline LSTM adversarial learning framework for identifying key-frames. In this paper, to better guide the network towards properly selecting video frames that can faithfully reconstruct the original video, we augment the baseline framework with an additional LSTM autoencoder, which learns in parallel a fixed-length representation of the entire original input sequence. This is exploited during training, where a novel loss term inspired by dictionary learning is added to the network optimization objectives, further biasing key-frame selection towards video frames which are collectively able to recreate the original video. Empirical evaluation on two common public relevant datasets indicates highly favourable results.

OTE: Optimal Trustworthy EdgeAI solutions for smart cities

“…The research challenge includes the development of corresponding DNN modules that will perform video instance segmentation, in real-time. The state-of-theart approach is to employ Transformer [17] or CNN neural architectures [18] and to exploit adversarial learning strategies to compensate for occlusions or distortions [19] and/or employ novel training goals that augment regular supervised training with unsupervised [20] or adversarial objectives [5], in order to increase accuracy in potential use-cases. Another challenge is to accelerate those algorithms for on-board execution without sacrificing accuracy, e.g., by combining multitask training on auxiliary tasks (such as scene geometry extraction by unsupervised depth map estimation).…”