Our daily life is highly influenced by what we consume and see. Attracting and holding one's attention -- the definition of (visual) interestingness -- is essential. The rise of Large Multimodal Models (LMMs) trained on large-scale visual and textual data has demonstrated impressive capabilities. We explore these models' potential to understand to what extent the concepts of visual interestingness are captured and examine the alignment between human assessments and GPT-4o's, a leading LMM, predictions through comparative analysis. Our studies reveal partial alignment between humans and GPT-4o. It already captures the concept as best compared to state-of-the-art methods. Hence, this allows for the effective labeling of image pairs according to their (commonly) interestingness, which are used as training data to distill the knowledge into a learning-to-rank model. The insights pave the way for a deeper understanding of human interest.

How to be successful on a video platform with 720K hours of video uploaded daily? This is the key question for every Youtube content creator. Especially, because there is only a thumbnail image and a short caption to attract the users’ attention and convince them to click a video. We introduce the Youtube-Science-Channels dataset containing more than 19k Youtube thumbnails and captions along with meta attributes, such as likes, dislikes and views from 40 channels. A Youtube video’s success is defined as generating lots of views. Given this labeling, a computational model is built in order to predict the success of a video using solely its thumbnail and caption. Experimental results are presented on both, the individual videos as well as aggregated per Youtube channel. Our combined vision and natural language model can accurately predict success. In addition, we analyse the effect of using clickbait in video captions. We show that clickbait is positively correlated with a Youtube channel’s success but that other factors such as high quality animations are important as well.

Source separation involving mono-channel audio is a challenging problem, in particular for speech separation where source contributions overlap both in time and frequency. This task is of high interest for applications such as video conferencing. Recent progress in machine learning has shown that the combination of visual cues, coming from the video, can increase the source separation performance. Starting from a recently designed deep neural network, we assess its ability and robustness to separate the visible speakers’ speech from other interfering speeches or signals. We test it for different configuration of video recordings where the speaker’s face may not be fully visible. We also asses the performance of the network with respect to different sets of visual features from the speakers’ faces.

We present a novel method for summarizing raw, casually captured videos. The objective is to create a short summary that still conveys the story. It should thus be both, interesting and representative for the input video. Previous methods often used simplified assumptions and only optimized for one of these goals. Alternatively, they used handdefined objectives that were optimized sequentially by making consecutive hard decisions. This limits their use to a particular setting. Instead, we introduce a new method that (i) uses a supervised approach in order to learn the importance of global characteristics of a summary and (ii) jointly optimizes for multiple objectives and thus creates summaries that posses multiple properties of a good summary. Experiments on two challenging and very diverse datasets demonstrate the effectiveness of our method, where we outperform or match current state-of-the-art.

The goal of integrating drones into the civil airspace requires a technical system which robustly detects, tracks and finally avoids aerial objects. Electro-optical cameras have proven to be an adequate sensor to detect traffic, especially for smaller aircraft, gliders or paragliders. However the very challenging environmental conditions and image artifacts such as lens flares often result in a high number of false detections. Depending on the solar radiation lens flares are very common in aerial images and hard to distinguish from aerial objects on a collision course due to their similar size, shape, brightness and trajectories. In this paper we present an efficient method to detect lens flares within aerial images based on the position of the sun with respect to the observer. Using the date, time, position and attitude of the observer we predict the lens flare direction within the image. Once the direction is known the position, size and shape of the lens flares are extracted. Experiments show that our approach is able to compensate for errors in the parameters influencing the calculation of the lens flare direction. We further integrate the lens flare detection into an aerial object tracking framework. A detailed evaluation of the framework with and without lens flare filter shows that false tracks due to lens flares are successfully suppressed without degrading the overall tracking system performance.

We investigate human interest in photos. Based on our own and others' psychophysical experiments, we identify various cues for "interestingness", namely aesthetics, unusualness and general preferences. For the ranking of retrieved images, interestingness shows to be more appropriate than cues proposed earlier. Interestingness is correlated with what people believe they will remember. This is opposed to actual memorability, which is uncorrelated to both. We introduce a set of features computationally capturing the three main aspects of visual interestingness and build an interestingness predictor from them. Its performance is shown on three datasets with varying context, reflecting the prior knowledge of the viewers.

Interestingness is said to be the power of attracting or holding one's attention (because something is unusual or exciting, etc.). We, as humans, have the great capacity to direct our visual attention and judge the interestingness of a scene. Consider for example the image sequence in the figure on the right. The spider in front of the camera or the snow on the lens are examples of events that deviate from the context since they violate the expectations, and therefore are considered interesting. On the other hand, weather changes or a camera shift, do not raise human attention considerably, even though large regions of the image are influenced. In this work we firstly investigate what humans consider as "interesting" in image sequences. Secondly we propose a computer vision algorithm to automatically spot these interesting events. To this end, we integrate multiple cues inspired by cognitive concepts and discuss why and to what extent the automatic discovery of visual interestingness is possible.

Person detection is a challenging task in industrial environments which typically feature rapidly changing conditions of illumination and the presence of occluding objects and cluttered background. This paper proposes a series of algorithms for improving the robustness of person detection in such harsh industrial environments. Based on a state-of-the-art person detector, significant robustness and automation is achieved by introducing automatic ground plane estimation, confidence filtering, cross-camera correspondence estimation and multi-camera fusion. Detailed experiments made on an industrial dataset that captures an automotive assembly process show the stepwise improvement when combining the above mentioned techniques in a fully unsupervised manner.

Temporal consistency is a strong cue in continuous data streams and especially in videos. We exploit this concept and encode temporal relations between consecutive frames using discriminative slow feature analysis. Activities are automatically segmented and represented in a hierarchical coarse to fine structure. Simultaneously, they are modeled in a generative manner, in order to analyze unseen data. This analysis supports the detection of previously learned activities and of abnormal, novel patterns. Our technique is purely data-driven and feature-independent. Experiments validate the approach in several contexts, such as traffic flow analysis and the monitoring of human behavior. The results are competitive with the state-of-the-art in all cases. © 2011. The copyright of this document resides with its authors.

We propose an efficient method for object localization and 3D pose estimation. A two-step approach is used. In the first step, a pose estimator is evaluated in the input images in order to estimate potential object locations and poses. These candidates are then validated, in the second step, by the corresponding pose-specific classifier. The result is a detection approach that avoids the inherent and expensive cost of testing the complete set of specific classifiers over the entire image. A further speedup is achieved by feature sharing. Features are computed only once and are then used for evaluating the pose estimator and all specific classifiers. The proposed method has been validated on two public datasets for the problem of detecting of cars under several views. The results show that the proposed approach yields high detection rates while keeping efficiency. © 2011. The copyright of this document resides with its authors.

Many object classes are primarily defined by their functions. However, this fact has been left largely unexploited by visual object categorization or detection systems. We propose a method to learn an affordance detector. It identifies locations in the 3d space which “support” the particular function. Our novel approach “imagines” an actor performing an action typical for the target object class, instead of relying purely on the visual object appearance. So, function is handled as a cue complementary to appearance, rather than being a consideration after appearance-based detection. Experimental results are given for the functional category “sitting”. Such affordance is tested on a 3d representation of the scene, as can be realistically obtained through SfM or depth cameras. In contrast to appearance-based object detectors, affordance detection requires only very few training examples and generalizes very well to other sittable objects like benches or sofas when trained on a few chairs.

Objects are usually embedded into context. Visual context has been successfully used in object detection tasks, however, it is often ignored in object tracking. We propose a method to learn supporters which are, be it only temporally, useful for determining the position of the object of interest. Our approach exploits the General Hough Transform strategy. It couples the supporters with the target and naturally distinguishes between strongly and weakly coupled motions. By this, the position of an object can be estimated even when it is not seen directly (e.g., fully occluded or outside of the image region) or when it changes its appearance quickly and significantly. Experiments show substantial improvements in model-free tracking as well as in the tracking of “virtual” points, e.g., in medical applications.

We propose a data-driven, hierarchical approach for the analysis of human actions in visual scenes. In particular, we focus on the task of in-house assisted living. In such scenarios the environment and the setting may vary considerably which limits the performance of methods with pre-trained models. Therefore our model of normality is established in a completely unsupervised manner and is updated automatically for scene-specific adaptation. The hierarchical representation on both an appearance and an action level paves the way for semantic interpretation. Furthermore we show that the model is suitable for coupled tracking and abnormality detection on different hierarchical stages. As the experiments show, our approach, simple yet effective, yields stable results, e.g. the detection of a fall, without any human interaction.

Automatic and real-time identification of unusual incidents is important for event detection and alarm systems. In today's camera surveillance solutions video streams are displayed on-screen for human operators, e.g. in large multi-screen control centers. This in turn requires the attention of operators for unusual events and urgent response.

We propose a method to recognize the traffic scene in front of a moving vehicle with respect to the road topology and the existence of objects. To this end, we use a two-stage system, where the first stage abstracts from the underlying image by means of a rough super-pixel segmentation of the scene. In a second stage, this meta representation is then used to construct a feature set for a classifier that is able to distinguish between different road types as well as detect the existence of commonly encountered objects, such as cars or pedestrian crossings. We show that by relying on an intermediate stage, we can effectively abstract from any peculiarities of the underlying image data due to e.g. color abberrations. The method is tested on two long, challenging urban data sets, covering both day light and dusk conditions. Compared to a state-of-the-art descriptor, we show improved classification performance, especially for object classes. © 2009. The copyright of this document resides with its authors.

In this paper we present an adaptive but robust object detector for static cameras by introducing classifier grids. Instead of using a sliding window for object detection we propose to train a separate classifier for each image location, obtaining a very specific object detector with a low false alarm rate. For each classifier corresponding to a grid element we estimate two generative representations in parallel, one describing the object's class and one describing the background. These are combined in order to obtain a discriminative model. To enable to adapt to changing environments these classifiers are learned on-line (i.e., boosting). Continuously learning (24 hours a day, 7 days a week) requires a stable system. In our method this is ensured by a fixed object representation while updating only the representation of the background. We demonstrate the stability in a long-term experiment by running the system for a whole week, which shows a stable performance over time. In addition, we compare the proposed approach to state-of-the-art methods in the field of person and car detection. In both cases we obtain competitive results.

The required amount of labeled training data for object detection and classification is a major drawback of current methods. Combining labeled and unlabeled data via semi-supervised learning holds the promise to ease the tedious and time consuming labeling effort. This paper presents a novel semi-supervised learning method which combines the power of learned similarity functions and classifiers. The approach capable of exploiting both labeled and unlabeled data is formulated in a boosting framework. One classifier (the learned similarity) serves as a prior which is steadily improved via training a second classifier on labeled and unlabeled samples. We demonstrate the approach on challenging computer vision applications. First, we show how we can train a classifier using only a few labeled samples and many unlabeled data. Second, we improve (specialize) a state-of-the-art detector by using labeled and unlabeled data.

On-line boosting allows to adapt a trained classifier to changing environmental conditions or to use sequentially available training data. Yet, two important problems in the on-line boosting training remain unsolved: (i) classifier evaluation speed optimization and, (ii) automatic classifier complexity estimation. In this paper we show how the on-line boosting can be combined with Waldpsilas sequential decision theory to solve both of the problems. The properties of the proposed on-line WaldBoost algorithm are demonstrated on a visual tracking problem. The complexity of the classifier is changing dynamically depending on the difficulty of the problem. On average, a speedup of a factor of 5-10 is achieved compared to the non-sequential on-line boosting.

We treat tracking as a matching problem of detected key-points between successive frames. The novelty of this paper is to learn classifier-based keypoint descriptions allowing to incorporate background information. Contrary to existing approaches, we are able to start tracking of the object from scratch requiring no off-line training phase before tracking. The tracker is initialized by a region of interest in the first frame. Afterwards an on-line boosting technique is used for learning descriptions of detected keypoints lying within the region of interest. New frames provide new samples for updating the classifiers which increases their stability. A simple mechanism incorporates temporal information for selecting stable features. In order to ensure correct updates a verification step based on estimating homographies using RANSAC is performed. The approach can be used for real-time applications since on-line updating and evaluating classifiers can be done efficiently.

A major shortcoming of discriminative recognition and detection methods is their noise sensitivity, both during training and recognition. This may lead to very sensitive and brittle recognition systems focusing on irrelevant information. This paper proposes a method that selects generative and discriminative features. In particular, we boost classical Haar-like features and use the same features to approximate a generative model (i.e., eigenimages). A modified error function for boosting ensures that only features are selected that show a good discrimination and reconstruction. This allows a robust feature selection using boosting. Thus, we can handle problems where discriminant classifiers fail while still retaining the discriminative power. Our experiments show that we can significantly improve the recognition performance when learning from noisy data. Moreover, the feature type used allows efficient recognition and reconstruction.

The automatic creation of 3D models of urban spaces has become a very active field of research. This has been inspired by recent applications in the location-awareness on the Internet, as demonstrated in maps.live.com and similar websites. The level of automation in creating 3D city models has increased considerably, and has benefited from an increase in the redundancy of the source imagery, namely digital aerial photography. In this paper we argue that the next big step forward is to replace photographic texture by an interpretation of what the texture describes, and to achieve this fully automatically. One calls the result "semantic knowledge". For example we want to know that a certain part of the image is a car, a person, a building, a tree, a shrub, a window, a door, instead of just a collection of 3D points or triangles with a superimposed photographic texture. We investigate object recognition methods to make this next big step. We demonstrate an early result of using the on-line variant of a Boosting algorithm to indeed detect cars in aerial digital imagery to a satisfactory and useful level of completeness. And we show that we can use this semantic knowledge to produce improved orthophotos. We expect that also the 3D models will be improved by the knowledge of cars.

This paper demonstrates how to reduce the hand labeling effort considerably by 3D information in an object detection task. In particular, we demonstrate how an efficient car detector for aerial images with minimal hand labeling effort can be build. We use an on-line boosting algorithm to incrementally improve the detection results. Initially, we train the classifier with a single positive (car) example, randomly drawn from a fixed number of given samples. When applying this detector to an image we obtain many false positive detections. We use information from a stereo matcher to detect some of these false positives (e.g. detected cars on a facade) and feed back this information to the classifier as negative updates. This improves the detector considerably, thus reducing the number of false positives. We show that we obtain similar results to hand labeling by iteratively applying this strategy. The performance of our algorithm is demonstrated on digital aerial images of urban environments.

Very recently tracking was approached using classification techniques such as support vector machines. The object to be tracked is discriminated by a classifier from the background. In a similar spirit we propose a novel on-line AdaBoost feature selection algorithm for tracking. The distinct advantage of our method is its capability of on-line training. This allows to adapt the classifier while tracking the object. Therefore appearance changes of the object (e.g. out of plane rotations, illumination changes) are handled quite naturally. Moreover, depending on the background the algorithm selects the most discriminating features for tracking resulting in stable tracking results. By using fast computable features (e.g. Haar-like wavelets, orientation histograms, local binary patterns) the algorithm runs in real-time. We demonstrate the performance of the algorithm on several (publically available) video sequences. 1

Boosting has become very popular in computer vision, showing impressive performance in detection and recognition tasks. Mainly off-line training methods have been used, which implies that all training data has to be a priori given; training and usage of the classifier are separate steps. Training the classifier on-line and incrementally as new data becomes available has several advantages and opens new areas of application for boosting in computer vision. In this paper we propose a novel on-line AdaBoost feature selection method. In conjunction with efficient feature extraction methods the method is real time capable. We demonstrate the multifariousness of the method on such diverse tasks as learning complex background models, visual tracking and object detection. All approaches benefit significantly by the on-line training.

In this paper we present the building blocks of our semi-automatic annotation tool which supports multi-modal and multi-level annotation of meetings. The main focus is on the proper design and functionality of the modules for recognizing meeting actions. The key features, identity and position of the speakers, are provided by different modalities (audio and video). Three audio algorithms (voice activity detection, speaker identification and direction of arrival) and three video algorithms (detection, tracking and identification) form the low-level feature extraction components. Low-level features are automatically merged and the recognized actions are proposed to the user by visualizing them. The annotation labels are related but not limited to events during meetings. The user can finally confirm or if necessary, modify the suggestion, and then store the actions into a database