Luc Van Gool

Bio: Luc Van Gool is an academic researcher from Katholieke Universiteit Leuven. The author has contributed to research in topics: Computer science & Object detection. The author has an hindex of 133, co-authored 1307 publications receiving 107743 citations. Previous affiliations of Luc Van Gool include Microsoft & ETH Zurich.

Configural and featural processing during face perception: A new stimulus set

Abstract: A new stimulus set of 60 male-face stimuli in seven in-depth orientations was developed. The set can be used in research on configural versus featural mechanisms of face processing. Configural, or holistic, changes are produced by changing the global form of the face, whereas featural, or part-based, changes are attained by altering the local form of internal facial features. For each face in the set, there is one other face that differs only by its global form and one other face that differs only by its internal features. In all faces, extrafacial cues have been eliminated or standardized. The stimulus set also contains a color-coded division of each face in areas of interest, which is useful for eye movement research on face scanning strategies. We report a matching experiment with upright and inverted face pairs that demonstrates that the face stimulus set is indeed useful for research on configural and featural face perception. The stimulus set may be downloaded from the Psychonomic Society's archive (brm.psychonomic-journals.org/content/supplemental) or from our Web site (http://ppw.kuleuven.be/labexppsy/newSite/resources).

Light, shadows and surface characteristics : the multispectral Portable Light Dome

Abstract: A multispectral, multidirectional, portable and dome-shaped acquisition system is developed within the framework of the research projects RICH (KU Leuven) and EES (RMAH, Brussels) in collaboration with the ESAT–VISICS research group (KU Leuven). The multispectral Portable Light Dome (MS PLD) consists of a hemispherical structure, an overhead camera and LEDs emitting in five parts of the electromagnetic spectrum regularly covering the dome’s inside surface. With the associated software solution, virtual relighting and enhancements can be applied in a real-time, interactive manner. The system extracts genuine 3D and shading information based on a photometric stereo algorithm. This innovative approach allows for instantaneous alternations between the computations in the infrared, red, green, blue and ultraviolet spectra. The MS PLD system has been tested for research ranging from medieval manuscript illuminations to ancient Egyptian artefacts. Preliminary results have shown that it documents and measures the 3D surface structure of objects, re-visualises underdrawings, faded pigments and inscriptions, and examines the MS results in combination with the actual relief characteristics of the physical object. Newly developed features are reflection maps and histograms, analytic visualisations of the reflection properties of all separate LEDs or selected areas. In its capacity as imaging technology, the system acts as a tool for the analysis of surface materials (e.g. identification of blue pigments, gold and metallic surfaces). Besides offering support in answering questions of attribution and monitoring changes and decay of materials, the PLD also contributes to the identification of materials, all essential factors when making decisions in the conservation protocol.

3D CNNs with Adaptive Temporal Feature Resolutions

Abstract: While state-of-the-art 3D Convolutional Neural Networks (CNN) achieve very good results on action recognition datasets, they are computationally very expensive and require many GFLOPs. While the GFLOPs of a 3D CNN can be decreased by reducing the temporal feature resolution within the network, there is no setting that is optimal for all input clips. In this work, we, therefore, introduce a differentiable Similarity Guided Sampling (SGS) module, which can be plugged into any existing 3D CNN architecture. SGS empowers 3D CNNs by learning the similarity of temporal features and grouping similar features together. As a result, the temporal feature resolution is not anymore static but it varies for each input video clip. By integrating SGS as an additional layer within current 3D CNNs, we can convert them into much more efficient 3D CNNs with adaptive temporal feature resolutions (ATFR). Our evaluations show that the proposed module improves the state-of-the-art by reducing the computational cost (GFLOPs)by half while preserving or even improving the accuracy. We evaluate our module by adding it to multiple state-of-the-art 3D CNNs on various datasets such as Kinetics-600, Kinetics-400, mini-Kinetics, Something-Something V2, UCF101, and HMDB51

A Survey on Deep Learning Technique for Video Segmentation.

Abstract: Video segmentation, i.e., partitioning video frames into multiple segments or objects, plays a critical role in a broad range of practical applications, e.g., visual effect assistance in movie, scene understanding in autonomous driving, and virtual background creation in video conferencing, to name a few. Recently, due to the renaissance of connectionism in computer vision, there has been an influx of numerous deep learning based approaches that have been dedicated to video segmentation and delivered compelling performance. In this survey, we comprehensively review two basic lines of research in this area, i.e., generic object segmentation (of unknown categories) in videos and video semantic segmentation, by introducing their respective task settings, background concepts, perceived need, development history, and main challenges. We also provide a detailed overview of representative literature on both methods and datasets. Additionally, we present quantitative performance comparisons of the reviewed methods on benchmark datasets. At last, we point out a set of unsolved open issues in this field, and suggest possible opportunities for further research.

Monocular tracking with a mixture of view-dependent learned models

Abstract: This paper considers the problem of monocular human body tracking using learned models. We propose to learn the joint probability distribution of appearance and body pose using a mixture of view-dependent models. In such a way the multimodal and nonlinear relationships can be captured reliably. We formulate inference algorithms that are based on generative models while exploiting the advantages of a learned model when compared to the traditionally used geometric body models. Given static images or sequences, body poses and bounding box locations are inferred using silhouette based image descriptors. Prior information about likely body poses and a motion model are taken into account. We consider analytical computations and Monte-Carlo techniques, as well as a combination of both. In a Rao-Blackwellised particle filter, the tracking problem is partitioned into a part that is solved analytically, and a part that is solved with particle filtering. Tracking results are reported for human locomotion

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.