Expressive Piano Performance Rendering from Unpaired Data

Recent advances in data-driven expressive performance rendering have enabled automatic models to reproduce the characteristics and the variability of human performances of musical compositions. However, these models need to be trained with aligned pairs of scores and performances and they rely notably on score-specific markings, which limits their scope of application. This work tackles the piano performance rendering task in a low-informed setting by only considering the score note information and without aligned data.
The proposed model relies on an adversarial training where the basic score notes properties are modified in order to reproduce the expressive qualities contained in a dataset of real performances. First results for unaligned score-to-performance rendering are presented through a conducted listening test. While the interpretation quality is not on par with highly-supervised methods and human renditions, our method shows promising results for transferring realistic expressivity into scores.

DDSP-Piano: A Neural Sound Synthesizer Informed by Instrument Knowledge

Instrument sound synthesis using deep neural networks has received numerous improvements over the last couple of years. Among them, the Differentiable Digital Signal Processing (DDSP) framework has modernized the spectral modeling paradigm by including signal-based synthesizers and effects into fully differentiable architectures. The present work extends the applications of DDSP to the task of polyphonic sound synthesis, with the proposal of a differentiable piano synthesizer conditioned on MIDI inputs. The model architecture is motivated by high-level acoustic modeling knowledge of the instrument, which, along with the sound structure priors inherent to the DDSP components, makes for a lightweight, interpretable, and realistic-sounding piano model. A subjective listening test has revealed that the proposed approach achieves better sound quality than a state-of-the-art neural-based piano synthesizer, but physical-modeling-based models still hold the best quality. Leveraging its interpretability and modularity, a qualitative analysis of the model behavior was also conducted: it highlights where additional modeling knowledge and optimization procedures could be inserted in order to improve the synthesis quality and the manipulation of sound properties. Eventually, the proposed differentiable synthesizer can be further used with other deep learning models for alternative musical tasks handling polyphonic audio and symbolic data.

Eliciting and Annotating Emotion in Virtual Spaces

We propose an online methodology where moment-to-moment affect annotations are gathered while exploring and visually interacting with virtual environments. For this task we developed an application to support this methodology, targeting both a VR and a desktop experience, and conducted a study to evaluate these two media of display. Results show that in terms of usability, both experiences were perceived equally positive. Presence was rated significantly higher for the VR experience, while participant ratings indicated a tendency for medium distraction during the annotation process. Additionally, effects between the architectural design elements were identified with perceived pleasure. The strengths and limitations of the proposed approach are highlighted to ground further work in gathering affect data in immersive and interactive media within the context of architectural appraisal. 


Knowing Your Annotator: Rapidly Testing the Reliability of Affect Annotation

The laborious and costly nature of affect annotation is a key detrimental factor for obtaining large scale corpora with valid and reliable affect labels. Motivated by the lack of tools that can effectively determine an annotator’s reliability, this paper proposes general quality assurance (QA) tests for real-time continuous annotation tasks. Assuming that the annotation tasks rely on stimuli with audiovisual components, such as videos, we propose and evaluate two QA tests: a visual and an auditory QA test. We validate the QA tool across 20 annotators that are asked to go through the test followed by a lengthy task of annotating the engagement of gameplay videos. Our findings suggest that the proposed QA tool reveals, unsurprisingly, that trained annotators are more reliable than the best of untrained crowdworkers we could employ. Importantly, the QA tool introduced can predict effectively the reliability of an affect annotator with 80% accuracy, thereby, saving on resources, effort and cost, and maximizing the reliability of labels solicited in affective corpora. The introduced QA tool is available and accessible through the PAGAN annotation platform.

Proceedings of the 3rd International Workshop on Learning to Quantify (LQ 2023)

The 3rd International Workshop on Learning to Quantify (LQ 2023 – https: // was held in Torino, IT, on September 18, 2023, as a satellite workshop of the European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML/PKDD 2023). While the 1st edition of the workshop (LQ 2021 – https://cikmlq2021. had to be an entirely online event, LQ 2023 (like the 2nd edi- tion LQ 2022 – was a hybrid event, with presentations given in-presence, and both in-presence attendees and remote attendees. The workshop was the second part (Sep 18 afternoon) of a full-day event, whose first part (Sep 18 morning) consisted of a tutorial on Learning to Quantify presented by Alejandro Moreo and Fabrizio Sebastiani. The LQ 2023 workshop consisted of the presentations of seven contributed papers, and a final collective discussion on the open problems of learning to quantify and on future initiatives.

The present volume contains five of the seven contributed papers that were accepted for presentation at the workshop (the authors of the other two papers decided not to have their paper included in the proceedings). Each contributed paper was submitted as a response to the call for papers, was reviewed by at least three members of the international program commit- tee, and was revised by the authors so as to take into account the feedback provided by the reviewers.

GreekPolitics: Sentiment Analysis on Greek Politically Charged Tweets

The rapid growth of on-line social media platforms has rendered opinion mining/sentiment analysis a critical area of research. This paper focuses on analyzing Twitter posts (tweets), written in the Greek language and politically charged in content. This is a rather underexplored topic, due to the inadequacy of publicly available annotated datasets. Thus, we present and release GreekPolitics: a dataset of Greek tweets with politically charged content, annotated for four different sentiments: polarity, figurativeness, aggressiveness and bias. GreekPolitics has been evaluated comprehensively using state-of-the-art Deep Neural Networks (DNNs) and data augmentation methods. This paper details the dataset, the evaluation process and the experimental results.

Quantifying the knowledge in Deep Neural Networks: an overview

Deep Neural Networks (DNNs) have proven to be extremely effective at learning a wide range of tasks. Due to their complexity and frequently inexplicable internal state, DNNs are difficult to analyze: their black-box nature makes it challenging for humans to comprehend their internal behavior. Several attempts to interpret their operation have been made during the last decade, but analyzing deep neural models from the perspective of the knowledge encoded in their layers is a very promising research direction, which has barely been touched upon. Such a research approach could provide a more accurate insight into a DNN model, its internal state, learning progress, and knowledge
storage capabilities. The purpose of this survey is two-fold: a) to review the concept of DNN knowledge quantification and highlight it as an important near-future challenge, as well as b) to provide a brief account of the scant existing methods attempting to actually quantify DNN knowledge. Although a few such algorithms have been proposed, this is an emerging topic still under investigation.

Political Tweet Sentiment Analysis For Public Opinion Polling

Public opinion measurement through polling is a classical political analysis task, e.g. for predicting national and local election results. However, polls are expensive to run and their results may be biased primarily due to improper population sampling. In this paper, we propose two innovative methods for employing tweet sentiment analysis’ results for public opinion polling. Our first method utilizes merely the tweet sentiment analysis’ results outperforming a plethora of well-recognised methods. In addition, we introduce a novel hybrid way to estimate electorally results from both public opinion polls and tweets. This method enables more accurate, frequent and inexpensive public opinion estimation and used for estimating the result of the 2023 Greek national election. Our method managed to achieve lower deviation than the conventional public opinion polls from the actual election’s results, introducing new possibilities for public opinion estimation using social media platforms.

Towards Human Society-inspired Decentralized DNN Inference

In human societies, individuals make their own decisions and they may select if and who may influence it, by e.g., consulting with people of their acquaintance or experts of a field. At a societal level, the overall knowledge is preserved and enhanced by individual person empowerment, where complicated consensus protocols have been developed over time in the form of societal mechanisms to assess, weight, combine and isolate individual people opinions. In distributed machine learning environments however, individual AI agents are merely part of a system where decisions are made in a centralized and aggregated fashion or require a fixed network topology, a practice prone to security risks and collaboration is nearly absent. For instance, Byzantine Failures may tamper both the training and inference stage of individual AI agents, leading to significantly reduced overall system performance. Inspired by societal practices, we propose a decentralized inference strategy where each individual agent is empowered to make their own decisions, by exchanging and aggregating information with other agents in their network. To this end, a ”Quality of Inference” consensus protocol (QoI) is proposed, forming a single commonly accepted inference rule applied by every individual agent. The overall system knowledge and decisions on specific manners can thereby be stored by all individual agents in a decentralized fashion, employing e.g., blockchain technology. Our experiments in classification tasks indicate that the proposed approach forms a secure decentralized inference framework, that prevents adversaries at tampering the overall process and achieves comparable performance with centralized decision aggregation methods.

Deep Reinforcement Learning with semi-expert distillation for autonomous UAV cinematography

Unmanned Aerial Vehicles (UAVs, or drones) have revolutionized modern media production. Being rapidly deployable “flying cameras”, they can easily capture aesthetically pleasing aerial footage of static or moving filming targets/subjects. Current approaches rely either on manual UAV/gimbal control by human experts or on a combination of complex computer vision algorithms and hardware configurations for automating the flight+flying process. This paper explores an efficient Deep Reinforcement Learning (DRL) alternative, which implicitly merges the target detection and path planning steps into a single algorithm. To achieve this, a baseline DRL approach is augmented with a novel policy distillation component, which transfers knowledge from a suitable, semi-expert Model Predictive Control (MPC) controller into the DRL agent. Thus, the latter is able to autonomously execute a specific UAV cinematography task with purely visual input. Unlike the MPC controller, the proposed DRL agent does not need to know the 3D world position of the filming target during inference. Experiments conducted in a photorealistic simulator showcase superior performance and training speed compared to the baseline agent while surpassing the MPC controller in terms of visual occlusion avoidance.

Knowledge Distillation-driven Communication Framework for Neural Networks: Enabling Efficient Student-Teacher Interactions

This paper presents a novel framework for facilitating communication and knowledge exchange among neural networks, leveraging the roles of both students and teachers. In our proposed framework, each node represents a neural network, capable of acting as either a student or a teacher. When new data is introduced and a network has not been trained on it, the node assumes the role of a student, initiating a communication process. The student node communicates with potential teachers, identifying those networks that have already been trained on the incoming data. Subsequently, the student node employs knowledge distillation techniques to learn from the teachers and gain insights from their accumulated knowledge. This approach enables efficient and effective knowledge transfer within the neural network ecosystem, enhancing learning capabilities and fostering collaboration among diverse networks. Experimental results demonstrate the efficacy of our framework in improving overall network performance and knowledge utilization.

Prompting Visual-Language Models for Dynamic Facial Expression Recognition

This paper presents a novel visual-language model called DFER-CLIP, which is based on the CLIP model and designed for in-the-wild Dynamic Facial Expression Recognition (DFER). Specifically, the proposed DFER-CLIP consists of a visual part and a textual part. For the visual part, based on the CLIP image encoder, a temporal model consisting of several Transformer encoders is introduced for extracting temporal facial expression features, and the final feature embedding is obtained as a learnable “class” token. For the textual part, we use as inputs textual descriptions of the facial behaviour that is related to the classes (facial expressions) that we are interested in recognising — those descriptions are generated using large language models, like ChatGPT. This, in contrast to works that use only the class names and more accurately captures the relationship between them. Alongside the textual description, we introduce a learnable token which helps the model learn relevant context information for each expression during training. Extensive experiments demonstrate the effectiveness of the proposed method and show that our DFER-CLIP also achieves state-of-the-art results compared with the current supervised DFER methods on the DFEW, FERV39k, and MAFW benchmarks.

MaskCon: Masked Contrastive Learning for Coarse-Labelled Dataset

Deep learning has achieved great success in recent years with the aid of advanced neural network structures and large-scale human-annotated datasets. However, it is often costly and difficult to accurately and efficiently annotate large-scale datasets, especially for some specialized domains where fine-grained labels are required. In this setting, coarse labels are much easier to acquire as they do not require expert knowledge. In this work, we propose a contrastive learning method, called masked contrastive learning (MaskCon) to address the under-explored problem setting, where we learn with a coarse-labelled dataset in order to address a finer labelling problem. More specifically, within the contrastive learning framework, for each sample our method generates soft-labels with the aid of coarse labels against other samples and another augmented view of the sample in question. By contrast to self-supervised contrastive learning where only the sample’s augmentations are considered hard positives, and in supervised contrastive learning where only samples with the same coarse labels are considered hard positives, we propose soft labels based on sample distances, that are masked by the coarse labels. This allows us to utilize both inter-sample relations and coarse labels. We demonstrate that our method can obtain as special cases many existing state-of-the-art works and that it provides tighter bounds on the generalization error. Experimentally, our method achieves significant improvement over the current state-of-the-art in various datasets, including CIFAR10, CIFAR100, ImageNet-1K, Standford Online Products and Stanford Cars196 datasets.

Self-Supervised Representation Learning with Cross-Context Learning between Global and Hypercolumn Features

Whilst contrastive learning yields powerful representations by matching different augmented views of the same instance, it lacks the ability to capture the similarities between different instances. One popular way to address this limitation is by learning global features (after the global pooling) to capture inter-instance relationships based on knowledge distillation, where the global features of the teacher are used to guide the learning of the global features of the student. Inspired by cross-modality learning, we extend this existing framework that only learns from global features by encouraging the global features and intermediate layer features to learn from each other. This leads to our novel self-supervised framework: cross-context learning between global and hypercolumn features (CGH), that enforces the consistency of instance relations between lowand high-level semantics. Specifically, we stack the intermediate feature maps to construct a “hypercolumn” representation so that we can measure instance relations using two contexts (hypercolumn and global feature) separately, and then use the relations of one context to guide the learning of the other. This cross-context learning allows the model to learn from the differences between the two contexts. The experimental results on linear classification and downstream tasks show that our method outperforms the state-of-the-art methods.

StyleMask: Disentangling the Style Space of StyleGAN2 for Neural Face Reenactment

In this paper we address the problem of neural face reenactment, where, given a pair of a source and a target facial image, we need to transfer the target’s pose (defined as the head pose and its facial expressions) to the source image, by preserving at the same time the source’s identity characteristics (e.g., facial shape, hair style, etc), even in the challenging case where the source and the target faces belong to different identities. In doing so, we address some of the limitations of the state-of-the-art works, namely, a) that they depend on paired training data (i.e., source and target faces have the same identity), b) that they rely on labeled data during inference, and c) that they do not preserve identity in large head pose changes. More specifically, we propose a framework that, using unpaired randomly generated facial images, learns to disentangle the identity characteristics of the face from its pose by incorporating the recently introduced style space S of StyleGAN2, a latent representation space that exhibits remarkable disentanglement properties. By capitalizing on this, we learn to successfully mix a pair of source and target style codes using supervision from a 3D model. The resulting latent code, that is subsequently used for reenactment, consists of latent units corresponding to the facial pose of the target only and of units corresponding to the identity of the source only, leading to notable improvement in the reenactment performance compared to recent state-of-the-art methods. In comparison to state of the art, we quantitatively and qualitatively show that the proposed method produces higher quality results even on extreme pose variations. Finally, we report results on real images by first embedding them on the latent space of the pretrained generator.

HyperReenact: One-Shot Reenactment via Jointly Learning to Refine and Retarget Faces

In this paper, we present our method for neural face reenactment, called HyperReenact, that aims to generate realistic talking head images of a source identity, driven by a target facial pose. Existing state-of-the-art face reenactment methods train controllable generative models that learn to synthesize realistic facial images, yet producing reenacted faces that are prone to significant visual artifacts, especially under the challenging condition of extreme head pose changes, or requiring expensive few-shot fine-tuning to better preserve the source identity characteristics. We propose to address these limitations by leveraging the photorealistic generation ability and the disentangled properties of a pretrained StyleGAN2 generator, by first inverting the real images into its latent space and then using a hypernetwork to perform:(i) refinement of the source identity characteristics and (ii) facial pose re-targeting, eliminating this way the dependence on external editing methods that typically produce artifacts. Our method operates under the one-shot setting (ie, using a single source frame) and allows for cross-subject reenactment, without requiring any subject-specific fine-tuning. We compare our method both quantitatively and qualitatively against several state-of-the-art techniques on the standard benchmarks of VoxCeleb1 and VoxCeleb2, demonstrating the superiority of our approach in producing artifact-free images, exhibiting remarkable robustness even under extreme head pose changes.

JGNN: Graph Neural Networks on Native Java

We introduce JGNN, an open source Java library to define, train, and run Graph Neural Networks (GNNs) under limited resources. The library is cross-platform and implements memory-efficient machine learning components without external dependencies. Model definition is simplified by parsing Python-like expressions, including interoperable dense and sparse matrix operations and inline parameter definitions. GNN models can be deployed on smart devices and trained on local data.

Temporal Normalization in Attentive Key-frame Extraction for Deep Neural Video Summarization

Attention-based neural architectures have consistently demonstrated superior performance over Long Short-Term Memory (LSTM) Deep Neural Networks (DNNs) in tasks such as key-frame extraction for video summarization. However, existing approaches mostly rely on rather shallow Transformer DNNs. This paper revisits the issue of model depth and proposes DATS: a deep attentive architecture for supervised video summarization that meaningfully exploits skip connections. Additionally, a novel per-layer temporal normalization algorithm is proposed that yields improved test accuracy. Finally, the model’s noisy output is rectified in an innovative post-processing step. Experiments conducted on two common, publicly available benchmark datasets showcase performance superior to competing state-of-the-art video summarization methods, both supervised and unsupervised.


Text-to-Motion Retrieval: Towards Joint Understanding of Human Motion Data and Natural Language

Due to recent advances in pose-estimation methods, human motion can be extracted from a common video in the form of 3D skeleton sequences. Despite wonderful application opportunities, effective and efficient content-based access to large volumes of such spatio-temporal skeleton data still remains a challenging problem. In this paper, we propose a novel content-based text-to-motion retrieval task, which aims at retrieving relevant motions based on a specified natural-language textual description. To define baselines for this uncharted task, we employ the BERT and CLIP language representations to encode the text modality and successful spatio-temporal models to encode the motion modality. We additionally introduce our transformer-based approach, called Motion Transformer (MoT), which employs divided space-time attention to effectively aggregate the different skeleton joints in space and time. Inspired by the recent progress in text-to-image/video matching, we experiment with two widely-adopted metric-learning loss functions. Finally, we set up a common evaluation protocol by defining qualitative metrics for assessing the quality of the retrieved motions, targeting the two recently-introduced KIT Motion-Language and HumanML3D datasets. The code for reproducing our results is available here:

Escaping local minima in deep reinforcement learning for video summarization

State-of-the-art deep neural unsupervised video summarization methods mostly fall under the adversarial reconstruction framework. This employs a Generative Adversarial Network (GAN) structure and Long Short-Term Memory (LSTM) auto-encoders during its training stage. The typical result is a selector LSTM that sequentially receives video frame representations and outputs corresponding scalar importance factors, which are then used to select key-frames. This basic approach has been augmented with an additional Deep Reinforcement Learning (DRL) agent, trained using the Discriminator’s output as a reward, which learns to optimize the selector’s outputs. However, local minima are a well-known problem in DRL. Thus, this paper presents a novel regularizer for escaping local loss minima, in order to improve unsupervised key-frame extraction. It is an additive loss term employed during a second training phase, that rewards the difference of the neural agent’s parameters from those of a previously found good solution. Thus, it encourages the training process to explore more aggressively the parameter space in order to discover a better local loss minimum. Evaluation performed on two public datasets shows considerable increases over the baseline and against the state-of-the-art.

Cross-Forgery Analysis of Vision Transformers and CNNs for Deepfake Image Detection

Deepfake Generation Techniques are evolving at a rapid pace, making it possible to create realistic manipulated images and videos and endangering the serenity of modern society. The continual emergence of new and varied techniques brings with it a further problem to be faced, namely the ability of deepfake detection models to update themselves promptly in order to be able to identify manipulations carried out using even the most recent methods. This is an extremely complex problem to solve, as training a model requires large amounts of data, which are difficult to obtain if the deepfake generation method is too recent. Moreover, continuously retraining a network would be unfeasible. In this paper, we ask ourselves if, among the various deep learning techniques, there is one that is able to generalise the concept of deepfake to such an extent that it does not remain tied to one or more specific deepfake generation methods used in the training set. We compared a Vision Transformer with an EfficientNetV2 on a cross-forgery context based on the ForgeryNet dataset. From our experiments, It emerges that EfficientNetV2 has a greater tendency to specialize often obtaining better results on training methods while Vision Transformers exhibit a superior generalization ability that makes them more competent even on images generated with new methodologies.