Adrien Deliège

Anthony Cioppa, Silvio Giancola, Vladimir Somers et al. · pku

The SoccerNet 2023 challenges were the third annual video understanding challenges organized by the SoccerNet team. For this third edition, the challenges were composed of seven vision-based tasks split into three main themes. The first theme, broadcast video understanding, is composed of three high-level tasks related to describing events occurring in the video broadcasts: (1) action spotting, focusing on retrieving all timestamps related to global actions in soccer, (2) ball action spotting, focusing on retrieving all timestamps related to the soccer ball change of state, and (3) dense video captioning, focusing on describing the broadcast with natural language and anchored timestamps. The second theme, field understanding, relates to the single task of (4) camera calibration, focusing on retrieving the intrinsic and extrinsic camera parameters from images. The third and last theme, player understanding, is composed of three low-level tasks related to extracting information about the players: (5) re-identification, focusing on retrieving the same players across multiple views, (6) multiple object tracking, focusing on tracking players and the ball through unedited video streams, and (7) jersey number recognition, focusing on recognizing the jersey number of players from tracklets. Compared to the previous editions of the SoccerNet challenges, tasks (2-3-7) are novel, including new annotations and data, task (4) was enhanced with more data and annotations, and task (6) now focuses on end-to-end approaches. More information on the tasks, challenges, and leaderboards are available on https://www.soccer-net.org. Baselines and development kits can be found on https://github.com/SoccerNet.

Anthony Cioppa, Silvio Giancola, Adrien Deliege et al.

Tracking objects in soccer videos is extremely important to gather both player and team statistics, whether it is to estimate the total distance run, the ball possession or the team formation. Video processing can help automating the extraction of those information, without the need of any invasive sensor, hence applicable to any team on any stadium. Yet, the availability of datasets to train learnable models and benchmarks to evaluate methods on a common testbed is very limited. In this work, we propose a novel dataset for multiple object tracking composed of 200 sequences of 30s each, representative of challenging soccer scenarios, and a complete 45-minutes half-time for long-term tracking. The dataset is fully annotated with bounding boxes and tracklet IDs, enabling the training of MOT baselines in the soccer domain and a full benchmarking of those methods on our segregated challenge sets. Our analysis shows that multiple player, referee and ball tracking in soccer videos is far from being solved, with several improvement required in case of fast motion or in scenarios of severe occlusion.

Silvio Giancola, Anthony Cioppa, Adrien Deliège et al.

The SoccerNet 2022 challenges were the second annual video understanding challenges organized by the SoccerNet team. In 2022, the challenges were composed of 6 vision-based tasks: (1) action spotting, focusing on retrieving action timestamps in long untrimmed videos, (2) replay grounding, focusing on retrieving the live moment of an action shown in a replay, (3) pitch localization, focusing on detecting line and goal part elements, (4) camera calibration, dedicated to retrieving the intrinsic and extrinsic camera parameters, (5) player re-identification, focusing on retrieving the same players across multiple views, and (6) multiple object tracking, focusing on tracking players and the ball through unedited video streams. Compared to last year's challenges, tasks (1-2) had their evaluation metrics redefined to consider tighter temporal accuracies, and tasks (3-6) were novel, including their underlying data and annotations. More information on the tasks, challenges and leaderboards are available on https://www.soccer-net.org. Baselines and development kits are available on https://github.com/SoccerNet.

Anthony Cioppa, Silvio Giancola, Vladimir Somers et al.

The SoccerNet 2024 challenges represent the fourth annual video understanding challenges organized by the SoccerNet team. These challenges aim to advance research across multiple themes in football, including broadcast video understanding, field understanding, and player understanding. This year, the challenges encompass four vision-based tasks. (1) Ball Action Spotting, focusing on precisely localizing when and which soccer actions related to the ball occur, (2) Dense Video Captioning, focusing on describing the broadcast with natural language and anchored timestamps, (3) Multi-View Foul Recognition, a novel task focusing on analyzing multiple viewpoints of a potential foul incident to classify whether a foul occurred and assess its severity, (4) Game State Reconstruction, another novel task focusing on reconstructing the game state from broadcast videos onto a 2D top-view map of the field. Detailed information about the tasks, challenges, and leaderboards can be found at https://www.soccer-net.org, with baselines and development kits available at https://github.com/SoccerNet.

Maxim Henry, Adrien Deliège, Sébastien Piérard et al.

Training high-capacity vision models from scratch requires substantial computational resources. To improve training efficiency of a wide target model, existing growth methods often assume the availability of narrower models, obscuring the true computational cost of the entire pipeline. We propose an efficient training protocol, RBDC, that builds wide models by coupling in a parameter-free block-diagonal way narrower, independently trained models in a recursive way. This allows a flexible allocation of the training budget available across all the models involved. Evaluated with vision transformers (DeiT) and convolutional networks (ResNet) on ImageNet, our RBDC training protocol shows a much better efficiency than models trained from scratch with the standard protocol, yielding 30% FLOPs reduction at similar test accuracies. It also achieves higher performances at same training FLOPs than training protocols from the model growth literature. Finally, we show that our models can serve as better backbones than their original counterparts for downstream object detection and instance segmentation tasks.

Adrien Deliège, Anthony Cioppa, Silvio Giancola et al.

Understanding broadcast videos is a challenging task in computer vision, as it requires generic reasoning capabilities to appreciate the content offered by the video editing. In this work, we propose SoccerNet-v2, a novel large-scale corpus of manual annotations for the SoccerNet video dataset, along with open challenges to encourage more research in soccer understanding and broadcast production. Specifically, we release around 300k annotations within SoccerNet's 500 untrimmed broadcast soccer videos. We extend current tasks in the realm of soccer to include action spotting, camera shot segmentation with boundary detection, and we define a novel replay grounding task. For each task, we provide and discuss benchmark results, reproducible with our open-source adapted implementations of the most relevant works in the field. SoccerNet-v2 is presented to the broader research community to help push computer vision closer to automatic solutions for more general video understanding and production purposes.

Jan Held, Renaud Vandeghen, Abdullah Hamdi et al.

Recent advances in radiance field reconstruction, such as 3D Gaussian Splatting (3DGS), have achieved high-quality novel view synthesis and fast rendering by representing scenes with compositions of Gaussian primitives. However, 3D Gaussians present several limitations for scene reconstruction. Accurately capturing hard edges is challenging without significantly increasing the number of Gaussians, creating a large memory footprint. Moreover, they struggle to represent flat surfaces, as they are diffused in space. Without hand-crafted regularizers, they tend to disperse irregularly around the actual surface. To circumvent these issues, we introduce a novel method, named 3D Convex Splatting (3DCS), which leverages 3D smooth convexes as primitives for modeling geometrically-meaningful radiance fields from multi-view images. Smooth convex shapes offer greater flexibility than Gaussians, allowing for a better representation of 3D scenes with hard edges and dense volumes using fewer primitives. Powered by our efficient CUDA-based rasterizer, 3DCS achieves superior performance over 3DGS on benchmarks such as Mip-NeRF360, Tanks and Temples, and Deep Blending. Specifically, our method attains an improvement of up to 0.81 in PSNR and 0.026 in LPIPS compared to 3DGS while maintaining high rendering speeds and reducing the number of required primitives. Our results highlight the potential of 3D Convex Splatting to become the new standard for high-quality scene reconstruction and novel view synthesis. Project page: convexsplatting.github.io.

Jan Held, Renaud Vandeghen, Adrien Deliege et al.

The field of computer graphics was revolutionized by models such as Neural Radiance Fields and 3D Gaussian Splatting, displacing triangles as the dominant representation for photogrammetry. In this paper, we argue for a triangle comeback. We develop a differentiable renderer that directly optimizes triangles via end-to-end gradients. We achieve this by rendering each triangle as differentiable splats, combining the efficiency of triangles with the adaptive density of representations based on independent primitives. Compared to popular 2D and 3D Gaussian Splatting methods, our approach achieves higher visual fidelity, faster convergence, and increased rendering throughput. On the Mip-NeRF360 dataset, our method outperforms concurrent non-volumetric primitives in visual fidelity and achieves higher perceptual quality than the state-of-the-art Zip-NeRF on indoor scenes. Triangles are simple, compatible with standard graphics stacks and GPU hardware, and highly efficient: for the \textit{Garden} scene, we achieve over 2,400 FPS at 1280x720 resolution using an off-the-shelf mesh renderer. These results highlight the efficiency and effectiveness of triangle-based representations for high-quality novel view synthesis. Triangles bring us closer to mesh-based optimization by combining classical computer graphics with modern differentiable rendering frameworks. The project page is https://trianglesplatting.github.io/

Anastasia Zakharova, Thierry Bouwmans, Anthony Cioppa et al.

Event detection tasks in videos, the most important aspect of video surveillance, aim to detect events either at the pixel-level, frame-level, or clip-level. Plenty of methods intended for event detection in different environments, for various applications, and within different acquisition techniques were introduced. Naturally, the attempts were made as well to classify these algorithms in terms of detection of performance or in terms of real-time abilities. Nevertheless, the lack of a large-scale dataset as well as rigorous performance evaluation methods have biased such comparisons as well as the development of the methods. Given the diversity of existing approaches, we believe it is essential for researchers to position their work within such a rich landscape. Thus, we propose a rigorous framework for developing new methods in event detection for videos. Specifically, this framework is based on three main pillars: datasets, performance evaluation, and scenarios for deploying methods.

Sébastien Piérard, Anaïs Halin, Anthony Cioppa et al.

Ranking entities such as algorithms, devices, methods, or models based on their performances, while accounting for application-specific preferences, is a challenge. To address this challenge, we establish the foundations of a universal theory for performance-based ranking. First, we introduce a rigorous framework built on top of both the probability and order theories. Our new framework encompasses the elements necessary to (1) manipulate performances as mathematical objects, (2) express which performances are worse than or equivalent to others, (3) model tasks through a variable called satisfaction, (4) consider properties of the evaluation, (5) define scores, and (6) specify application-specific preferences through a variable called importance. On top of this framework, we propose the first axiomatic definition of performance orderings and performance-based rankings. Then, we introduce a universal parametric family of scores, called ranking scores, that can be used to establish rankings satisfying our axioms, while considering application-specific preferences. Finally, we show, in the case of two-class classification, that the family of ranking scores encompasses well-known performance scores, including the accuracy, the true positive rate (recall, sensitivity), the true negative rate (specificity), the positive predictive value (precision), and F1. However, we also show that some other scores commonly used to compare classifiers are unsuitable to derive performance orderings satisfying the axioms.

Sébastien Piérard, Anaïs Halin, Anthony Cioppa et al.

In the computer vision and machine learning communities, as well as in many other research domains, rigorous evaluation of any new method, including classifiers, is essential. One key component of the evaluation process is the ability to compare and rank methods. However, ranking classifiers and accurately comparing their performances, especially when taking application-specific preferences into account, remains challenging. For instance, commonly used evaluation tools like Receiver Operating Characteristic (ROC) and Precision/Recall (PR) spaces display performances based on two scores. Hence, they are inherently limited in their ability to compare classifiers across a broader range of scores and lack the capability to establish a clear ranking among classifiers. In this paper, we present a novel versatile tool, named the Tile, that organizes an infinity of ranking scores in a single 2D map for two-class classifiers, including common evaluation scores such as the accuracy, the true positive rate, the positive predictive value, Jaccard's coefficient, and all F-beta scores. Furthermore, we study the properties of the underlying ranking scores, such as the influence of the priors or the correspondences with the ROC space, and depict how to characterize any other score by comparing them to the Tile. Overall, we demonstrate that the Tile is a powerful tool that effectively captures all the rankings in a single visualization and allows interpreting them.

Sébastien Piérard, Adrien Deliège, Marc Van Droogenbroeck

Ranking methods or models based on their performance is of prime importance but is tricky because performance is fundamentally multidimensional. In the case of classification, precision and recall are scores with probabilistic interpretations that are both important to consider and complementary. The rankings induced by these two scores are often in partial contradiction. In practice, therefore, it is extremely useful to establish a compromise between the two views to obtain a single, global ranking. Over the last fifty years or so,it has been proposed to take a weighted harmonic mean, known as the F-score, F-measure, or $F_β$. Generally speaking, by averaging basic scores, we obtain a score that is intermediate in terms of values. However, there is no guarantee that these scores lead to meaningful rankings and no guarantee that the rankings are good tradeoffs between these base scores. Given the ubiquity of $F_β$ scores in the literature, some clarification is in order. Concretely: (1) We establish that $F_β$-induced rankings are meaningful and define a shortest path between precision- and recall-induced rankings. (2) We frame the problem of finding a tradeoff between two scores as an optimization problem expressed with Kendall rank correlations. We show that $F_1$ and its skew-insensitive version are far from being optimal in that regard. (3) We provide theoretical tools and a closed-form expression to find the optimal value for $β$ for any distribution or set of performances, and we illustrate their use on six case studies.

Sébastien Piérard, Adrien Deliège, Anaïs Halin et al.

Frequently, multiple entities (methods, algorithms, procedures, solutions, etc.) can be developed for a common task and applied across various domains that differ in the distribution of scenarios encountered. For example, in computer vision, the input data provided to image analysis methods depend on the type of sensor used, its location, and the scene content. However, a crucial difficulty remains: can we predict which entities will perform best in a new domain based on assessments on known domains, without having to carry out new and costly evaluations? This paper presents an original methodology to address this question, in a leave-one-domain-out fashion, for various application-specific preferences. We illustrate its use with 30 strategies to predict the rankings of 40 entities (unsupervised background subtraction methods) on 53 domains (videos).

Adrien Deliège, Maxime Istasse, Ashwani Kumar et al.

In the framework of convolutional neural networks, downsampling is often performed with an average-pooling, where all the activations are treated equally, or with a max-pooling operation that only retains an element with maximum activation while discarding the others. Both of these operations are restrictive and have previously been shown to be sub-optimal. To address this issue, a novel pooling scheme, named\emph{ ordinal pooling}, is introduced in this work. Ordinal pooling rearranges all the elements of a pooling region in a sequence and assigns a different weight to each element based upon its order in the sequence. These weights are used to compute the pooling operation as a weighted sum of the rearranged elements of the pooling region. They are learned via a standard gradient-based training, allowing to learn a behavior anywhere in the spectrum of average-pooling to max-pooling in a differentiable manner. Our experiments suggest that it is advantageous for the networks to perform different types of pooling operations within a pooling layer and that a hybrid behavior between average- and max-pooling is often beneficial. More importantly, they also demonstrate that ordinal pooling leads to consistent improvements in the accuracy over average- or max-pooling operations while speeding up the training and alleviating the issue of the choice of the pooling operations and activation functions to be used in the networks. In particular, ordinal pooling mainly helps on lightweight or quantized deep learning architectures, as typically considered e.g. for embedded applications.

Adrien Deliège, Anthony Cioppa, Marc Van Droogenbroeck

Supervised image classification problems rely on training data assumed to have been correctly annotated; this assumption underpins most works in the field of deep learning. In consequence, during its training, a network is forced to match the label provided by the annotator and is not given the flexibility to choose an alternative to inconsistencies that it might be able to detect. Therefore, erroneously labeled training images may end up ``correctly'' classified in classes which they do not actually belong to. This may reduce the performances of the network and thus incite to build more complex networks without even checking the quality of the training data. In this work, we question the reliability of the annotated datasets. For that purpose, we introduce the notion of ghost loss, which can be seen as a regular loss that is zeroed out for some predicted values in a deterministic way and that allows the network to choose an alternative to the given label without being penalized. After a proof of concept experiment, we use the ghost loss principle to detect confusing images and erroneously labeled images in well-known training datasets (MNIST, Fashion-MNIST, SVHN, CIFAR10) and we provide a new tool, called sanity matrix, for summarizing these confusions.

Anthony Cioppa, Adrien Deliège, Floriane Magera et al.

Soccer broadcast video understanding has been drawing a lot of attention in recent years within data scientists and industrial companies. This is mainly due to the lucrative potential unlocked by effective deep learning techniques developed in the field of computer vision. In this work, we focus on the topic of camera calibration and on its current limitations for the scientific community. More precisely, we tackle the absence of a large-scale calibration dataset and of a public calibration network trained on such a dataset. Specifically, we distill a powerful commercial calibration tool in a recent neural network architecture on the large-scale SoccerNet dataset, composed of untrimmed broadcast videos of 500 soccer games. We further release our distilled network, and leverage it to provide 3 ways of representing the calibration results along with player localization. Finally, we exploit those representations within the current best architecture for the action spotting task of SoccerNet-v2, and achieve new state-of-the-art performances.

Anthony Cioppa, Adrien Deliège, Noor Ul Huda et al.

Monitoring the occupancy of public sports facilities is essential to assess their use and to motivate their construction in new places. In the case of a football field, the area to cover is large, thus several regular cameras should be used, which makes the setup expensive and complex. As an alternative, we developed a system that detects players from a unique cheap and wide-angle fisheye camera assisted by a single narrow-angle thermal camera. In this work, we train a network in a knowledge distillation approach in which the student and the teacher have different modalities and a different view of the same scene. In particular, we design a custom data augmentation combined with a motion detection algorithm to handle the training in the region of the fisheye camera not covered by the thermal one. We show that our solution is effective in detecting players on the whole field filmed by the fisheye camera. We evaluate it quantitatively and qualitatively in the case of an online distillation, where the student detects players in real time while being continuously adapted to the latest video conditions.

Anthony Cioppa, Adrien Deliège, Silvio Giancola et al.

In video understanding, action spotting consists in temporally localizing human-induced events annotated with single timestamps. In this paper, we propose a novel loss function that specifically considers the temporal context naturally present around each action, rather than focusing on the single annotated frame to spot. We benchmark our loss on a large dataset of soccer videos, SoccerNet, and achieve an improvement of 12.8% over the baseline. We show the generalization capability of our loss for generic activity proposals and detection on ActivityNet, by spotting the beginning and the end of each activity. Furthermore, we provide an extended ablation study and display challenging cases for action spotting in soccer videos. Finally, we qualitatively illustrate how our loss induces a precise temporal understanding of actions and show how such semantic knowledge can be used for automatic highlights generation.

A. Deliege, A. Cioppa, M. Van Droogenbroeck

Neural networks designed for the task of classification have become a commodity in recent years. Many works target the development of more effective networks, which results in a complexification of their architectures with more layers, multiple sub-networks, or even the combination of multiple classifiers, but this often comes at the expense of producing uninterpretable black boxes. In this paper, we redesign a simple capsule network to enable it to synthesize class-representative samples, called prototypes, by replacing the last layer with a novel Hit-or-Miss layer. This layer contains activated vectors, called capsules, that we train to hit or miss a fixed target capsule by tailoring a specific centripetal loss function. This possibility allows to develop a data augmentation step combining information from the data space and the feature space, resulting in a hybrid data augmentation process. We show that our network, named HitNet, is able to reach better performances than those reproduced with the initial CapsNet on several datasets, while allowing to visualize the nature of the features extracted as deformations of the prototypes, which provides a direct insight into the feature representation learned by the network .

Adrien Deliège, Anthony Cioppa, Marc Van Droogenbroeck

Neural networks designed for the task of classification have become a commodity in recent years. Many works target the development of better networks, which results in a complexification of their architectures with more layers, multiple sub-networks, or even the combination of multiple classifiers. In this paper, we show how to redesign a simple network to reach excellent performances, which are better than the results reproduced with CapsNet on several datasets, by replacing a layer with a Hit-or-Miss layer. This layer contains activated vectors, called capsules, that we train to hit or miss a central capsule by tailoring a specific centripetal loss function. We also show how our network, named HitNet, is capable of synthesizing a representative sample of the images of a given class by including a reconstruction network. This possibility allows to develop a data augmentation step combining information from the data space and the feature space, resulting in a hybrid data augmentation process. In addition, we introduce the possibility for HitNet, to adopt an alternative to the true target when needed by using the new concept of ghost capsules, which is used here to detect potentially mislabeled images in the training data.