Call for Papers and Award
Timeline
The deadline for paper submission is Jul, 24th, 2024 (11:59PM, Pacific time) Jul, 19th, 2024 (11:59PM, Pacific time).
Decision to authors will be released on Aug, 19th, 2024.
The camera ready deadline is set to Aug, 25th, 2024 (11:59PM, Pacific time).
Submission instructions
Submission will follow the ECCV format (more detailed instructions will be published here, see the ECCV website for format instruction).
Submit your paper here: https://cmt3.research.microsoft.com/CVV2024/.
Topics
We invite authors to submit their works in one of the following (or affine) areas:
Quality assessment (QA) automation: As QA is a significant part of game development, the gaming industry is trying to automate it through CV. Typical QA tasks include detection of rendering artifacts, task difficulty control, character balancing, and more. Visual quality could also be assessed in terms of player advantage/disadvantage in a competitive environment.
Reinforcement and imitation learning: Autonomous agents are essential elements in videogames. We look forward to applications of recent breakthroughs in this highly cross-disciplinary field such as generalistic agents, sample efficient learning algorithms, decision transformers, continuous learning, collaborative agents, and more.
Human perception models: While game engines can control lighting or texture parameters, it is difficult to predict the end effects on players' visual understanding. Efficient yet accurate models of player perception substantially enhances game design process and player experience.
Human behavior models: Accurate modeling of human behaviors can add depth to videogame playing experiences on multiple aspects: realistic character animation, personalization, design of in-game tasks, character skills, and interface. Human behavior models can also find their applications in competitive gaming for coaching, training, play-style classification, and more.
Security, privacy, and prevention of toxicity: We call for CV/AI techniques to protect and keep playing experience healthy: detection/removal of verbal abuse, harassment, griefing, cheating, or improper in-game gestures. System approaches such as memory sniffing are beyond the workshop scope.
Language models and applications: Recent advances in language models and their multi-modality have potential to change experiences in videogames and their development. We welcome emerging ideas and their applications such as automatic commentary, non-player character (NPC) narrations, automatic summarization, and more.
Asset creation and management: Generative AI and multimodal language model will change how rendering assets are created and managed. One example is the generation of levels/worlds/animations from reference videos, using a library of curated assets. Multimodal language models can also significantly reduce overhead in managing rendering assets.
Agent-based frameworks: With in-game autonomous agents being more intelligent and generalistic, the way games are developed and enjoyed can fundamentally change. These agents can and will conduct more active roles of suggesting next steps, detecting problems, and even generating solutions. We welcome any visionary works on this emerging thought framework.
Datasets: Large datasets with videogames data are not easily found in the public domain. We want to sparkle discussion on copyright issues, data acquisition, preprocessing and synchronization - all topics that are already relevant and well known in the CV community, but assume a peculiar nuance in the context of videogames.
Computationally efficient methods: Computer vision techniques must be computationally efficient to add value to videogames and their applications. This includes realtime rendering problems such as global illumination, post-processing methods such as super sampling, and creation of large, parallelizable simulation systems for training and testing.
CV for animation and control: Infrared or time of flight cameras, among others, are devices that strongly influenced the development of natural human computer interfaces in the past years. CV is fundamental for the processing the data generated by these devices that are widely used to interact as videogame interfaces.
Serious games: All the aforementioned topics are relevant for simulated environments that go beyond the entertainment scope, including for instance systems for surgery training, post stroke cognitive rehabilitation, and so on.
Award
The best academic paper will be awarded with a first-class GPU provided by NVIDIA.