Search

Video activity localization aims at understanding the semantic content in long untrimmed videos and retrieving actions of interest. The retrieved action with its start and end locations can be used for highlight generation, temporal action detection, etc. Unfortunately, learning the exact boundary location of activities is highly challenging because temporal activities are continuous in time, and there are often no clear-cut transitions between actions. Moreover, the definition of the start and end of events is subjective, which may confuse the model. To alleviate the boundary ambiguity, we propose to study the video activity localization problem from a denoising perspective. Specifically, we propose an encoder-decoder model named DenoiseLoc. During training, a set of action spans is randomly generated from the ground truth with a controlled noise scale. Then we attempt to reverse this process by boundary denoising, allowing the localizer to predict activities with precise boundaries and resulting in faster convergence speed. Experiments show that DenoiseLoc advances %in several video activity understanding tasks. For example, we observe a gain of +12.36% average mAP on QV-Highlights dataset and +1.64% mAP@0.5 on THUMOS'14 dataset over the baseline. Moreover, DenoiseLoc achieves state-of-the-art performance on TACoS and MAD datasets, but with much fewer predictions compared to other current methods.

Polynomial regression is a recurrent problem with a large number of applications. In computer vision it often appears in motion analysis. Whatever the application, standard methods for regression of polynomial models tend to deliver biased results when the input data is heavily contaminated by outliers. Moreover, the problem is even harder when outliers have strong structure. Departing from problem-tailored heuristics for robust estimation of parametric models, we explore deep convolutional neural networks. Our work aims to find a generic approach for training deep regression models without the explicit need of supervised annotation. We bypass the need for a tailored loss function on the regression parameters by attaching to our model a differentiable hard-wired decoder corresponding to the polynomial operation at hand. We demonstrate the value of our findings by comparing with standard robust regression methods. Furthermore, we demonstrate how to use such models for a real computer vision problem, i.e., video stabilization. The qualitative and quantitative experiments show that neural networks are able to learn robustness for general polynomial regression, with results that well overpass scores of traditional robust estimation methods., 18 pages, conference

International audience; We propose an original method for detecting and localizing anomalous motion patterns in videos from a camera view-based motion representation perspective. Anomalous motion should be taken in a broad sense, i.e., unexpected, abnormal, singular, irregular, or unusual motion. Identifying distinctive dynamic information at any time point and at any image location in a sequence of images is a key requirement in many situations and applications. The proposed method relies on so-called labeled affine flows (LAF) involving both affine velocity vectors and affine motion classes. At every pixel, a motion class is inferred from the affine motion model selected in a set of candidate models estimated over a collection of windows. Then, the image is subdivided in blocks where motion class histograms weighted by the affine motion vector magnitudes are computed. They are compared blockwise to histograms of normal behaviors with a dedicated distance. More specifically, we introduce the local outlier factor (LOF) to detect anomalous blocks. LOF is a local flexible measure of the relative density of data points in a feature space, here the space of LAF histograms. By thresholding the LOF value, we can detect an anomalous motion pattern in any block at any time instant of the video sequence. The threshold value is automatically set in each block by means of statistical arguments. We report comparative experiments on several real video datasets, demonstrating that our method is highly competitive for the intricate task of detecting different types of anomalous motion in videos. Specifically, we obtain very competitive results on all the tested datasets: 99.2% AUC for UMN, 82.8% AUC for UCSD, and 95.73% accuracy for PETS 2009, at the frame level.