Biblio

In recent years, developments in the field of computer vision have allowed deep learning-based techniques to surpass human-level performance. However, these advances have also culminated in the advent of adversarial machine learning techniques, capable of launching targeted image captioning attacks that easily fool deep learning models. Although attacks in the image domain are well studied, little work has been done in the video domain. In this paper, we show it is possible to extend prior attacks in the image domain to the video captioning task, without heavily affecting the video's playback quality. We demonstrate our attack against a state-of-the-art video captioning model, by extending a prior image captioning attack known as Show and Fool. To the best of our knowledge, this is the first successful method for targeted attacks against a video captioning model, which is able to inject 'subliminal' perturbations into the video stream, and force the model to output a chosen caption with up to 0.981 cosine similarity, achieving near-perfect similarity to chosen target captions.

This paper investigates several techniques that increase the accuracy of motion boundaries in estimated motion fields of a local dense estimation scheme. In particular, we examine two matching metrics, one is MSE in the image domain and the other one is a recently proposed multiresolution metric that has been shown to produce more accurate motion boundaries. We also examine several different edge-preserving filters. The edge-aware moving average filter, proposed in this paper, takes an input image and the result of an edge detection algorithm, and outputs an image that is smooth except at the detected edges. Compared to the adoption of edge-preserving filters, we find that matching metrics play a more important role in estimating accurate and compressible motion fields. Nevertheless, the proposed filter may provide further improvements in the accuracy of the motion boundaries. These findings can be very useful for a number of recently proposed scalable interactive video coding schemes.