Comprehensive difficulty in a game can be defined as the challenge presented by learning the game's system, encompassing fundamental rules, objectives, and parameters. Achieving a well-balanced difficulty level is crucial for a game's success. In the industry, rules and objectives are commonly designed using a Static Game Balance (SGB) system. This involves human designers handcrafting elements such as game mechanics and the core behavior of enemies. In the early stages of game production, particularly during prototyping without access to player data, it becomes challenging to objectively assess the difficulty of these manually designed elements. Limited research has addressed automatic difficulty assessment in the context of SGB, with most studies focusing on executive difficulty, which pertains to a player's motor skills, such as dexterity. However, the industry is in need of more automated software tools to optimize game production. In this paper, we propose a novel method for automatically measuring the comprehensive difficulty of game enemies. Our approach, owing to its generalizability and the standardized way of defining enemies using state machines, lever-ages the properties of graphs (ex. number of states, transitions, cyclomatic complexity, etc.) to establish a standardized model for comprehensive difficulty assessment in games. We present this model along with the results of an initial exploratory experiment, demonstrating the potential of our approach and its feasibility for integration as a plugin in game engines like Unreal Engine 5.
Deepfakes are hyper-realistic videos in which the faces are replaced, swapped, or forged using deep-learning models. This potent media manipulation techniques hold promise for applications across various domains. Yet, they also present a significant risk when employed for malicious intents like identity fraud, phishing, spreading false information, and executing scams. In this work, we propose a novel and improved Deepfake video detector that uses a Convolutional Vision Transformer (CViT2), which builds on the concepts of our previous work (CViT). The CViT architecture consists of two components: a Convolutional Neural Network that extracts learnable features, and a Vision Transformer that categorizes these learned features using an attention mechanism. We trained and evaluted our model on 5 datasets, namely Deepfake Detection Challenge Dataset (DFDC), FaceForensics++ (FF++), Celeb-DF v2, Deep-fakeTIMIT, and TrustedMedia. On the test sets unseen during training, we achieved an accuracy of 95%, 94.8%, 98.3% and 76.7% on the DFDC, FF++, Celeb-DF v2, and TIMIT datasets, respectively. In conclusion, our proposed Deepfake detector can be used in the battle against misinformation and other forensic use cases.
The representation of politics in media influences societal perceptions and attitudes. Video games, as a pervasive form of media, contribute significantly to this phenomenon. In this work, we explore political themes within video games by analyzing politically-themed games on game distribution plat-forms including Steam. We conducted a statistical examination of games with political context to identify patterns and use this as a basis to introduce a first taxonomy to categorize and better understand the interplay between politics and video games. This taxonomy offers a first framework for analyzing political content in games and also sets a foundation for future research in this field
The rapid advancement of generative artificial in-telligence (AI) has led to groundbreaking developments in large language models. As large language models generate textual sequences autoregressively, mitigating latency becomes imper-ative for providing a highly immersive interaction experience within a realtime conversation, for example, providing fast and accurate responses to users' questions. Current efforts focus on accelerating inference processes, yet often at the expense of model architecture alterations, leading to compromised quality. In this paper, we explore latency reduction in the case of speech-based conversational agents. We leverage mathematical functions based on Beam Search to analyze autoregressive textual sequences, enabling a nuanced evaluation of semantic quality during auditory interaction, for example, for use within interactive web podcasts. We implemented our concepts and used the software to evaluate the concepts within (1) an automated evaluation of 1000 question-answer pairs and (2) a user survey. The results show that the semantic quality of autoregressive textual sequences could be assessed successfully by our proposed mathematical terms.
Recently, the importance of having a representation of the users inside Virtual Reality (VR) by using avatars, which replicate the movements of real humans, has grown in several fields of application. Nowadays, developers can use many off-the-shelf devices for visualizations and tracking. Game engines like Unity3D and Unreal allow to easily integrate packages to manage head-mounted displays and other tracking devices. However, it is still difficult to combine different solutions and achieve a full representation of the human body, e.g., combining full-body and hand tracking or switching among different tracking modalities in an easy way. This paper describes IMMERSE, an open-source framework based on Unity3D, the XR Interaction toolkit, and an inverse kinematics solver. IMMERSE allows developers to insert a VR avatar in a virtual environment, animate it through different sources of 6DOF tracking measurements, which are also interchangeable during the simulation, and finally record the movements to animate other avatars or input to motion analysis techniques.
