We present a multi-temporal, multi-modal remote-sensing dataset for predicting how active wildfires will spread at a resolution of 24 hours. The dataset consists of 13607 images across 607 fire events in the United States from January 2018 to October 2021. For each fire event, the dataset contains a full time series of daily observations, containing detected active fires and variables related to fuel, topography and weather conditions. The dataset is challenging due to: a) its inputs being multi-temporal, b) the high number of 23 multi-modal input channels, c) highly imbalanced labels and d) noisy labels, due to smoke, clouds, and inaccuracies in the active fire detection. The underlying complexity of the physical processes adds to these challenges. Compared to existing public datasets in this area, WildfireSpreadTS allows for multi-temporal modeling of spreading wildfires, due to its time series structure. Furthermore, we provide additional input modalities and a high spatial resolution of 375m for the active fire maps. We publish this dataset to encourage further research on this important task with multi-temporal, noise-resistant or generative methods, uncertainty estimation or advanced optimization techniques that deal with the high-dimensional input space.

The Visible Infrared Imaging Radiometer Suite (VIIRS) onboard the Suomi National Polar-orbiting Partnership (S-NPP) satellite has been used for the early detection and daily monitoring of active wildfires. How to effectively segment the active fire pixels from VIIRS image time-series in a reliable manner remains a challenge because of the low precision associated with high recall using automatic methods. For active fire detection, multi-criteria thresholding is often applied to both low-resolution and mid-resolution Earth observation images. Deep learning approaches based on Convolutional Neural Networks are also well-studied on mid-resolution images. However, ConvNet-based approaches have poor performance on low-resolution images because of the coarse spatial features. On the other hand, the high temporal resolution of VIIRS images highlights the potential of using sequential models for active fire detection. Transformer networks, a recent deep learning architecture based on self-attention, offer hope as they have shown strong performance on image segmentation and sequential modelling tasks within computer vision. In this research, we propose a Transformer-based solution to segment active fire pixels from the VIIRS time-series. The solution feeds a time-series of tokenized pixels into a Transformer network to identify active fire pixels at each timestamp and achieves a significantly higher F1-Score than prior approaches for active fires within the study areas in California, New Mexico, and Oregon in the US, and in British Columbia and Alberta in Canada, as well as in Australia and Sweden.

The number of linear regions has been studied as a proxy of complexity for ReLU networks. However, the empirical success of network compression techniques like pruning and knowledge distillation, suggest that in the overparameterized setting, linear regions density might fail to capture the effective nonlinearity. In this work, we propose an efficient algorithm for discovering linear regions and use it to investigate the effectiveness of density in capturing the nonlinearity of trained VGGs and ResNets on CIFAR-10 and CIFAR-100. We contrast the results with a more principled nonlinearity measure based on function variation, highlighting the shortcomings of linear regions density. Furthermore, interestingly, our measure of nonlinearity clearly correlates with model-wise deep double descent, connecting reduced test error with reduced nonlinearity, and increased local similarity of linear regions.

This paper focuses on estimating probability distributions over the set of 3D ro- tations (SO(3)) using deep neural networks. Learning to regress models to the set of rotations is inherently difficult due to differences in topology between R^N and SO(3). We overcome this issue by using a neural network to out- put the parameters for a matrix Fisher distribution since these parameters are homeomorphic to R^9 . By using a negative log likelihood loss for this distri- bution we get a loss which is convex with respect to the network outputs. By optimizing this loss we improve state-of-the-art on several challenging applica- ble datasets, namely Pascal3D+, ModelNet10-SO(3). Our code is available at https://github.com/Davmo049/Publicproborientationestimationwith_matrix fisherdistributions

Predicting facial attributes from faces in the wild is very challenging due to pose and lighting variations in the real world. The key to this problem is to build proper feature representations to cope with these unfavourable conditions. Given the success of Convolutional Neural Network (CNN) in image classification, the high-level CNN feature, as an intuitive and reasonable choice, has been widely utilized for this problem. In this paper, however, we consider the mid-level CNN features as an alternative to the high-level ones for attribute prediction. This is based on the observation that face attributes are different: some of them are locally oriented while others are globally defined. Our investigations reveal that the mid-level deep representations outperform the prediction accuracy achieved by the (fine-tuned) high-level abstractions. We empirically demonstrate that the mid-level representations achieve state-of-the-art prediction performance on CelebA and LFWA datasets. Our investigations also show that by utilizing the mid-level representations one can employ a single deep network to achieve both face recognition and attribute prediction.

This paper addresses the problem of transferring CNNs pre-trained for face recognition to a face attribute prediction task. To transfer an off-the-shelf CNN to a novel task, a typical solution is to fine-tune the network towards the novel task. As demonstrated in the state-of-the-art face attribute prediction approach, fine-tuning the high-level CNN hidden layer by using labeled attribute data leads to significant performance improvements. In this paper, however, we tackle the same problem but through a different approach. Rather than using an end-to-end network, we select face descriptors from off-the-shelf hierarchical CNN representations for recognizing different attributes. Through such an adaptive representation selection, even without any fine-tuning, our results still outperform the state-of-the-art face attribute prediction approach on the latest large-scale dataset for an error rate reduction of more than 20%. Moreover, by using intensive empirical probes, we have identified several key factors that are significant for achieving promising face attribute prediction performance. These results attempt to gain and update our understandings of the nature of CNN features and how they can be better applied to the transferred novel tasks.

Evidence is mounting that Convolutional Networks (ConvNets) are the most effective representation learning method for visual recognition tasks. In the common scenario, a ConvNet is trained on a large labeled dataset (source) and the feed-forward units activation of the trained network, at a certain layer of the network, is used as a generic representation of an input image for a task with relatively smaller training set (target). Recent studies have shown this form of representation transfer to be suitable for a wide range of target visual recognition tasks. This paper introduces and investigates several factors affecting the transferability of such representations. It includes parameters for training of the source ConvNet such as its architecture, distribution of the training data, etc. and also the parameters of feature extraction such as layer of the trained ConvNet, dimensionality reduction, etc. Then, by optimizing these factors, we show that significant improvements can be achieved on various (17) visual recognition tasks. We further show that these visual recognition tasks can be categorically ordered based on their similarity to the source task such that a correlation between the performance of tasks and their similarity to the source task w.r.t. the proposed factors is observed.

This paper addresses the problem of Face Alignment for a single image. We show how an ensemble of regression trees can be used to estimate the face's landmark positions directly from a sparse subset of pixel intensities, achieving super-realtime performance with high quality predictions. We present a general framework based on gradient boosting for learning an ensemble of regression trees that optimizes the sum of square error loss and naturally handles missing or partially labelled data. We show how using appropriate priors exploiting the structure of image data helps with efficient feature selection. Different regularization strategies and its importance to combat overfitting are also investigated. In addition, we analyse the effect of the quantity of training data on the accuracy of the predictions and explore the effect of data augmentation using synthesized data.

This paper demonstrates a system for automatic detection of visual attention and identification of salient items at exhibitions (e.g. museum or an auction). The method is offline and is done on a video captured by a head mounted camera. Towards the estimation of attention, we define the notions of "saliency" and "interestingness" for an exhibition items. Our method is a combination of multiple state of the art techniques from different vision tasks such as tracking, image matching and retrieval. Many experiments are conducted to evaluate multiple aspects of our method. The method has proven to be robust to image blur, occlusion, truncation, and dimness. The experiments shows strong performance for the tasks of matching items, estimating focus frames and detecting salient and interesting items. This can be useful to the commercial vendors and museum curators and help them to understand which items are appealing more to the visitors.

This paper addresses the problem of human pose estimation, given images taken from multiple dynamic but calibrated cameras. We consider solving this task using a part-based model and focus on the part appearance component of such a model. We use a random forest classifier to capture the variation in appearance of body parts in 2D images. The result of these 2D part detectors are then aggregated across views to produce consistent 3D hypotheses for parts. We solve correspondences across views for mirror symmetric parts by introducing a latent variable. We evaluate our part detectors qualitatively and quantitatively on a dataset gathered from a professional football game.

For object classification and detection, the algorithm pipeline often involves classifying feature vectors extracted from image patches. Existing features such as HOG, fail to map the image patches into a space where a linear hyperplane is suitable for separating the classes, while many non-linear classification methods are too expensive for many tasks. We propose a sparse tree-based mapping method that learns a mapping of the feature vector to a space where a linear hyperplane can better separate negative and positive examples. The learned mapping function F(x) results in significant improvement for image patch classification with HOG and LBP-features over other feature mapping methods on VOC2007 and INRIAPerson datasets.

We consider the problem of automatically estimating the 3D pose of humans from images, taken from multiple calibrated views. We show that it is possible and tractable to extend the pictorial structures framework, popular for 2D pose estimation, to 3D. We discuss how to use this framework to impose view, skeleton, joint angle and intersection constraints in 3D. The 3D pictorial structures are evaluated on multiple view data from a professional football game. The evaluation is focused on computational tractability, but we also demonstrate how a simple 2D part detector can be plugged into the framework.

Many computer vision tasks such as object detection, pose estimation,and alignment are directly related to the estimation of correspondences overinstances of an object class. Other tasks such as image classification andverification if not completely solved can largely benefit from correspondenceestimation. This thesis presents practical approaches for tackling the corre-spondence estimation problem with an emphasis on deformable objects.Different methods presented in this thesis greatly vary in details but theyall use a combination of generative and discriminative modeling to estimatethe correspondences from input images in an efficient manner. While themethods described in this work are generic and can be applied to any object,two classes of objects of high importance namely human body and faces arethe subjects of our experimentations.When dealing with human body, we are mostly interested in estimating asparse set of landmarks – specifically we are interested in locating the bodyjoints. We use pictorial structures to model the articulation of the body partsgeneratively and learn efficient discriminative models to localize the parts inthe image. This is a common approach explored by many previous works. Wefurther extend this hybrid approach by introducing higher order terms to dealwith the double-counting problem and provide an algorithm for solving theresulting non-convex problem efficiently. In another work we explore the areaof multi-view pose estimation where we have multiple calibrated cameras andwe are interested in determining the pose of a person in 3D by aggregating2D information. This is done efficiently by discretizing the 3D search spaceand use the 3D pictorial structures model to perform the inference.In contrast to the human body, faces have a much more rigid structureand it is relatively easy to detect the major parts of the face such as eyes,nose and mouth, but performing dense correspondence estimation on facesunder various poses and lighting conditions is still challenging. In a first workwe deal with this variation by partitioning the face into multiple parts andlearning separate regressors for each part. In another work we take a fullydiscriminative approach and learn a global regressor from image to landmarksbut to deal with insufficiency of training data we augment it by a large numberof synthetic images. While we have shown great performance on the standardface datasets for performing correspondence estimation, in many scenariosthe RGB signal gets distorted as a result of poor lighting conditions andbecomes almost unusable. This problem is addressed in another work wherewe explore use of depth signal for dense correspondence estimation. Hereagain a hybrid generative/discriminative approach is used to perform accuratecorrespondence estimation in real-time.

We propose a system for the automatic segmentation of novelties from the background in scenarios where multiple images of the same environment are available e.g. obtained by wearable visual cameras. Our method finds the pixels in a query image corresponding to the underlying background environment by comparing it to reference images of the same scene. This is achieved despite the fact that all the images may have different viewpoints, significantly different illumination conditions and contain different objects - cars, people, bicycles, etc. - occluding the background. We estimate the probability of each pixel, in the query image, belonging to the background by computing its appearance inconsistency to the multiple reference images. We then, produce multiple segmentations of the query image using an iterated graph cuts algorithm, initializing from these estimated probabilities and consecutively combine these segmentations to come up with a final segmentation of the background. Detection of the background in turn highlights the novel pixels. We demonstrate the effectiveness of our approach on a challenging outdoors data set.

We propose a new method for face alignment with part-based modeling. This method is competitive in terms of precision with existing methods such as Active Appearance Models, but is more robust and has a superior generalization ability due to its part-based nature. A variation of the Histogram of Oriented Gradients descriptor is used to model the appearance of each part and the shape information is represented with a set of landmark points around the major facial features. Multiple linear regression models are learnt to estimate the position of the landmarks from the appearance of each part. We verify our algorithm with a set of experiments on human faces and these show the competitive performance of our method compared to existing methods.

This paper demonstrates a system for the automatic extraction of novelty in images captured from a small video camera attached to a subject's chest, replicating his visual perspective, while performing activities which are repeated daily. Novelty is detected when a (sub)sequence cannot be registered to previously stored sequences captured while performing the same daily activity. Sequence registration is performed by measuring appearance and geometric similarity of individual frames and exploiting the invariant temporal order of the activity. Experimental results demonstrate that this is a robust way to detect novelties induced by variations in the wearer's ego-motion such as stopping and talking to a person. This is an essentially new and generic way of automatically extracting information of interest to the camera wearer and can be used as input to a system for life logging or memory support.

We explore the benefits of using contextual features for head pose estimation in football games. Contextual features are derived from knowledge of the position of all players and combined with image based features derived from low-resolution footage. Using feature selection and combination techniques, we show that contextual features can aid head pose estimation in football games and potentially be an important complement to the image based features traditionally used.

In this paper we introduce a new kind of feature - the multi-local feature, so named as each one is a collection of local features, such as oriented edgels, in a very specific spatial arrangement. A multi-local feature has the ability to capture underlying constant shape properties of exemplars from an object class. Thus it is particularly suited to representing and detecting visual classes that lack distinctive local structures and are mainly defined by their global shape. We present algorithms to automatically learn an ensemble of these features to represent an object class from weakly labelled training images of that class, as well as procedures to detect these features efficiently in novel images. The power of multi-local features is demonstrated by using the ensemble in a simple voting scheme to perform object category detection on a standard database. Despite its simplicity, this scheme yields detection rates matching state-of-the-art object detection systems.

Many objects have smooth surfaces of a fairly uniform color, thereby exhibiting shading patterns that reveal information about its shape, an important clue to the nature of the object. This papers explores extracting this information from images, by creating shape detectors based on shading.

Multi-target tracking requires locating the targets and labeling their identities. The latter is a challenge when many targets, with indistinct appearances, frequently occlude one another, as in football and surveillance tracking. We present an approach to solving this labeling problem. When isolated, a target can be tracked and its identity maintained. While, if targets interact this is not always the case. This paper assumes a track graph exists, denoting when targets are isolated and describing how they interact. Measures of similarity between isolated tracks are defined. The goal is to associate the identities of the isolated tracks, by exploiting the graph constraints and similarity measures. We formulate this as a Bayesian network inference problem, allowing us to use standard message propagation to find the most probable set of paths in an efficient way. The high complexity inevitable in large problems is gracefully reduced by removing dependency links between tracks. We apply the method to a 10 min sequence of an international football game and compare results to ground truth.

Edge maps which are computed from textured scenes using existing methods based on local image analysis are not very meaningful. This is because edges at object boundaries are not differentiated from edges in texture. We introduce a real-time algorithm that overcomes this difficulty by computing the Dirichlet integral in a small image patch at different scales. These measurements are combined and interpreted in a probabilistic framework avoiding the need for a threshold. As a result the output of this algorithm can be utilised by a higher level process. Texture does not fit our model of a boundary edge thus its presence is detected by the probabilistic model as an outlier. Convincing results are shown on synthetic as well as images of the natural world. This algorithm is intended to be a fast preprocessing step for localising boundary edges and textured image regions.

Maximisation of cross correlation is a commonly used principle for intensity based object localization that gives a single estimate of location. However, to facilitate sequential inference (e.g. over time or scale) and to allow the representation of ambiguity, it is desirable to represent an entire probability distribution for object location. Although the cross correlation itself (or some function of it) has sometimes been treated as a probability distribution, this is not generally justifiable. Bayesian correlation achieves a consistent probabilistic treatment by combining several developments. The first is the interpretation of correlation matching functions in probabilistic terms, as observation likelihoods. Second, probability distributions of filter bank responses are learned from training examples. Inescapably, response learning also demands statistical modelling of background intensities, and there are links here with image coding and Independent Component Analysis. Lastly, multi scale processing is achieved in a Bayesian context by means of a new algorithm, layered sampling, for which asymptotic properties are derived.