Diverse video captioning through latent variable expansion

• A diverse captioning model of full convolution design is proposed. We develop a new evaluation metric to assess the sentence diversity. Our method achieves superior performance compared state-of-the-art benchmarks. Automatically describing video content with text description challenging but important task, which has been attracting lot attention in computer vision community. Previous works mainly strive for accuracy generated sentences, while ignoring sentences diversity, inconsistent human behavior. In this paper, we aim caption each multiple descriptions and propose novel framework. Concretely, given video, intermediate latent variables conventional encode-decode process are utilized as input conditional generative adversarial network (CGAN) purpose generating sentences. adopt different Convolutional Neural Networks (CNNs) our generator that produces conditioned on discriminator assesses quality Simultaneously, DCE designed captions. evaluate benchmark datasets, where it demonstrates its ability generate results against other methods.