ImageCLEF 2027

News

Motivation

ImageCLEF 2027 is organized as part of the CLEF Initiative Labs.

The target audience for ImageCLEF 2027 is mainly expected to be from the multimodal data annotation and retrieval community, from fields such as computer vision, image information retrieval and digital image processing. Due to the success and the specific nature of the medical tasks, a significant part of the audience will come from the medical informatics, machine learning and pattern recognition community.

Stay tuned with us for the latest information and updates by joining us on the ImageCLEF social media accounts: Twitter #imageclef, and Facebook @ImageClef.

ImageCLEF2027 schedule

Each of the tasks sets its own schedule, so please check the corresponding task webpage for specific dates. A (tentative) global schedule can be found below:

• TBD: Registration opens for all ImageCLEF tasks
• TBD: Development dataset released (depends on task)
• TBD: Test dataset released (depends on task)
• TBD: Registration closes for all ImageCLEF tasks
• TBD: Deadline for submitting participant runs
• TBD: Release of the processed results by the task organizers (depends on task)
• TBD: Submission of participant papers [CEUR-WS]
• TBD: Notification of acceptance
• TBD: CLEF 2027, TBD

The CLEF Conference

The CLEF 2027 conference will be hosted in TBD, on TBD.
ImageCLEF lab and all its tasks are part of the Conference and Labs of the Evaluation Forum: CLEF 2027. CLEF 2027 will be hosted in TBD, on TBD, and consists of an independent peer-reviewed workshops on a broad range of challenges in the fields of multilingual and multimodal information access evaluation, and a set of benchmarking activities carried in various labs designed to test different aspects of mono and cross-language information retrieval systems. More details about the conference can be found here. Also there is more information about the CLEF Initiative.

Programme of ImageCLEF at the CLEF 2027 Conference

Participant registration

The Tasks

ImageCLEF 2027 proposes 5 main tasks:

(11th edition) ImageCLEFmedical: Multimodal data can be used in different scenarios. For example, manual generation of the knowledge of medical images is a time-consuming process prone to human error. As this process requires assistance for the better and easier diagnoses of diseases that are susceptible to radiology screening, it is important that we better understand and refine automatic systems that aid in the broad task of radiology-image metadata generation. Thus, we proposed:

• (11th edition) Automatic Image Captioning challenges participants to detect radiological concepts, generate coherent image-level captions, and provide human-interpretable explanations for those captions in medical images.
• (5th edition) Synthetic Medical Images Created via GANs: investigates the security and privacy implications of generative models for medical data by challenging participants to attribute synthetic images to their generating architecture, generate realistic synthetic medical images, and detect adversarially perturbed medical images.
• (5th edition) Visual Question Answering: The MEDVQA-GI 2027 challenge advances clinically grounded visual question answering for GI endoscopy by evaluating accurate diagnosis-oriented answers together with safe, explainable, and medically justified multimodal reasoning, with a stronger emphasis on robustness under challenging clinical conditions.
• (3rd edition) MEDIQA-MAGIC: challenges participants to jointly reason over doctor–patient dialogues and dermatological images, segmenting the described skin lesions and answering closed-ended clinical questions supported by evidence from the dialogue or the image.

(4th edition) ImageCLEFtoPicto: This task is designed for individuals with language impairments who use pictograms as a communication aid. The main usage scenario involves converting text into a meaningful sequence of pictograms, facilitating communication between verbal individuals and AAC users.

(3rd edition) ImageCLEF MultimodalReasoning: Vision-Language Models (VLMs) excel in tasks combining vision and language, like image captioning and basic visual question answering. However, they often falter in deep logical reasoning and handling complex dependencies or hypothetical scenarios. This task aims to evaluate modern LLMs' reasoning abilities on intricate, multilingual inputs across diverse subjects.

(2nd edition) Deepfake Detection and Generation: challenges participants to both generate and detect audio, image and video deepfakes, jointly evaluating how realistic synthetic media can be made and how robust detection systems are against increasingly sophisticated forgeries.

(1st edition) CvTR QA: Multilingual Multimodal Chart and Tabular Question Answering challenges participants to extract, understand, and reason over structured information presented in charts and tables, across multiple languages and across visual, code-based, and textual representations.

Overview Paper

The Organising Committee

Overall coordination

• Bogdan Ionescu <bogdan.ionescu(at)upb.ro>, National University of Science and Technology Politehnica Bucharest, Romania
• Henning Müller <henning.mueller(at)hevs.ch>, University of Applied Sciences Western Switzerland, Sierre, Switzerland
• Cristian Stanciu <dan.stanciu1203(at)upb.ro>, National University of Science and Technology Politehnica Bucharest, Romania

Technical support

• Ivan Eggel <ivan.eggel(at)hevs.ch>, University of Applied Sciences Western Switzerland, Sierre, Switzerland
• Liviu-Daniel Ștefan <liviu_daniel.stefan(at)upb.ro>, National University of Science and Technology Politehnica Bucharest, Romania