Nude Detection in Video Using Bag-of-Visual-Features

Abstract

Abstract-The ability to filter improper content from multimedia sources based on visual content has important applications, since text-based filters are clearly insufficient against erroneous and/or malicious associations between text and actual content. In this paper, we investigate a method for detection of nudity in videos based on a bag-of-visual-features representation for frames and an associated voting scheme.Bag-of-Visual-Features (BoVF) approaches have been successfully applied to object recognition and scene classification, showing robustness to occlusion and also to the several kinds of variations that normally curse object detection methods. To the best of our knowledge, only two proposals in the literature use BoVF for nude detection in still images, and no other attempt has been made at applying BoVF for videos. Nevertheless, the results of our experiments show that this approach is indeed able to provide good recognition rates for nudity even at the frame level and with a relatively low sampling ratio. Also, the proposed voting scheme significantly enhances the recognition rates for video segments, achieving, in the best case, a value of 93.2% of correct classification, using a sampling ratio of 1/15 frames. Finally, a visual analysis of some particular cases indicates possible sources of misclassifications.

Content-based obscene video recognition by combining 3D spatiotemporal and motion-based features

“…In [14], Bag-of-Visual-Features was used for nudity detection in video files. The features used to build the vocabulary in this method were simply patches (gray-level values) around the interest points.…”

Adult Video Content Detection Using Machine Learning Techniques

“…By means of these techniques it is possible to identify and block adult content before it reaches the wrong audiences. A number of solutions exist for adult image detection [2,3,4,5,6], and fewer have been proposed to deal with video mainly extending the image-based methods [7,8,9] or using a more sophisticated set of features exploiting the temporal cues in video [10,11,12]. Taking into account that adult content detection is a video classification problem, a number of interesting and promising methods to be specifically applied for the mentioned purpose are reviewed [13,14,15,16,17,18].…”

“…The most typical means of identifying pornography are strongly based in skin detection [21,6,22]. In the case of videos, the most common approach is basically an extension of the image detection techniques [7,8,9] applied for a set of selected frames representative of the clip known as key frames. More novel techniques utilize the temporal cues as well to extract information such as motion patterns [10,11,12].…”

“…Even though SVMs are just employable for n-binary classification, they have high accuracy ratios and have the interesting property of being insensitive to the relative number of positive and negative examples in the training set [18]. Therefore they are widely used for image and video categorization problems [10,7,8]. Among the different kernels functions that can be used to draw the boundary between classes, the Gaussian (RBF) kernel is preferred for content based video genre classification [47] and adult content detection [18] having just the inconvenience of being sensible to the γ parameter.…”

“…Among the different kernels functions that can be used to draw the boundary between classes, the Gaussian (RBF) kernel is preferred for content based video genre classification [47] and adult content detection [18] having just the inconvenience of being sensible to the γ parameter. LibSVM tool is referenced by many visual content-based applications [48,46,8,49,31]. In an experiment conducted by [18], where a specific framework to extract a discriminative feature based on skin detection is also proposed, different SVM kernels are compared.…”