Call for Papers: 29th International Conference on Pattern Recognition 2027

The International Conference on Pattern Recognition Virtual (ICPRv) is the virtual extension of the flagship physical conference of the International Association for Pattern Recognition, the 29th edition of ICPR. Like its physical counterpart, this fully virtual conference encompasses a wide range of topics where Pattern Recognition methods are applied in fields including Computer Vision, Machine Learning, Image Processing, Speech and Natural Language Processing, and Sensor Pattern Processing.

The Inaugural ICPRv, to be held in 2027, offers an excellent platform for students, academics, and industry researchers to foster new ideas and collaborations. Unlike the face-to-face ICPR conferences, ICPRv will not have tracks and plenary sessions. Instead, all papers will be presented in two-hour thematic sessions of papers on the same topic. This will create sessions with a more focused and specialized workshop-like feel, with opportunities for discussion and interaction with the authors. Authors will identify their preferred thematic session topic when they submit their paper. Delegates can attend any and all of the live thematic sessions, which will also be recorded for the benefit of delegates wanting to attend simultaneous sessions or sessions at inconvenient times.

As well as the traditional ICPR conference topics listed at the end of this CFP, we encourage papers on truly novel research problems. Examples include (but not restricted to):

Universal Neuro-Symbolic Knowledge Engines: Foundation models still hallucinate and update knowledge poorly. New architectures should fuse LLMs with dynamic knowledge graphs to absorb rich documents, update world models without retraining, and enable provable reasoning over evolving scientific and enterprise data.
Omnimodal Embodied World Models: Current research still treats vision, audio, and text separately. The next step is to build omnimodal world models that integrate physics and asynchronous IoT signals, enabling robots to understand real-world causal rules and transfer smoothly from simulation to deployment.
Continuous-Time Fluid Architectures: Efficient AI still relies on discrete tokens and batching. The next frontier is continuous, asynchronous architectures that adapt compute to signal entropy, using little energy for simple data and deeper reasoning only for anomalies or complex patterns.
Mechanistic Fairness and Provable Unlearning: Current trustworthy AI methods remain largely post-hoc. The next step is to combine mechanistic interpretability and algorithmic fairness so models can isolate and remove specific biases or sensitive content, with provable guarantees of privacy and fairness for critical domains such as law and medicine.
Adversarial Co-Evolutionary Data Ecosystems: As generative AI advances, the generator–detector arms race is becoming unsustainable. A next step is closed-loop ecosystems where generators and forensic models co-evolve during training, embedding cryptographic provenance and robust invisible watermarks directly into generated outputs.
Causal Discovery Foundation Models: Causal discovery remains an emerging field. The next leap is foundation models that can design experiments, intervene in simulations, and infer causal equations from observational data, moving beyond prediction to uncover the physical or economic laws governing a system.
Hyper-Personalized Cognitive Digital Twins: Personalized systems remain separate from advanced LLM reasoning. The next step is privacy-preserving, edge-based cognitive twins that learn a user’s physiological, emotional, and behavioral patterns, anticipate complex needs, and keep sensitive biometric and behavioral data strictly on-device.

Efficient and scalable AI Embodied vision

Event-based cameras

Explainability and interpretability in pattern recognition

Few-shot and zero-shot learning

Generative models

Graph-based and Bayesian models

Human Computer Interaction

Image analysis and recognition

Image and video processing

Image and video synthesis and generation

Image detection and segmentation

Large Language Models (LLMs) and Vision Language Models (VLMs)

Low-level vision

Machine learning (inc. supervised, unsupervised, semi- and weakly-supervised, etc.)

Medical and biomedical imaging, cell microscopy

Motion and video analysis

Multimodal and multi-label learning

Natural language processing (NLP)

Neural networks

Object detection and recognition

Online, continual, and active learning

Optimization methods

Photogrammetry and remote sensing

Physics-based vision and shape-from-X

Representation learning

Scene analysis and understanding

Social signal processing

Text detection and recognition

Theory of computer vision and pattern recognition

Transfer learning

Transparency, fairness, accountability, privacy, ethics

Vision applications and systems

Vision, language, and reasoning

Call for Papers

Submission and Review:

Paper Format and Length:

Besides the novel research areas described above, ICPRv continues to welcome traditional ICPR topics as listed below: