International Journal of Computer Vision (IJCV) details the science and engineering of this rapidly growing field. Regular articles present major technical advances of broad general interest. Survey articles offer critical reviews of the state of the art and/or tutorial presentations of pertinent topics.

Coverage includes:

- Mathematical, physical and computational aspects of computer vision: image formation, processing, analysis, and interpretation; machine learning techniques; statistical approaches; sensors.
- Applications: image-based rendering, computer graphics, robotics, photo interpretation, image retrieval, video analysis and annotation, multi-media, and more.
- Connections with human perception: computational and architectural aspects of human vision.

The journal also features book reviews, position papers, editorials by leading scientific figures, as well as additional on-line material, such as still images, video sequences, data sets, and software.

Please note: the median time indicated below is computed over all the submitted manuscripts including the ones that are not put into the review pipeline at the onset of the review process. The typical time to first decision for manuscripts is approximately 96 days.

Special Issue on Computer Vision Approaches for Animal Tracking and Modeling 2025
提出日: 2025-09-20

Guest editors

    Urs Waldmann, Linköping University, Sweden
    Shangzhe Wu, University of Cambridge, United Kingdom
    Hedvig Kjellström, KTH Royal Institute of Technology, Sweden
    Albert Ali Salah, Utrecht University, Netherlands
    Isla Camille Duporge, Princeton University, NJ, USA

Many biological organisms have evolved to exhibit diverse behaviors, and understanding these behaviors is a fundamental goal of multiple disciplines including neuroscience, biology, animal husbandry, ecology, and animal conservation. These analyses require objective, repeatable, and scalable measurements of animal behaviors that are not possible with existing methodologies that leverage manual encoding from animal experts and specialists. Computer vision is having an impact across multiple disciplines by providing new tools for the detection, tracking, and analysis of animal behavior. The fourth edition of CV4Animals workshop has brought together this year experts across fields to stimulate this new field of computer-vision-based animal behavioral understanding.

This special issue will feature invited extended papers presented at the CV4Animals workshop in Nashville, Tennessee, USA in June 2025. This is an invite-only issue; only authors who have received an invitation are eligible to submit to this special issue. Appropriate submissions include, but are not limited to the following forms or a combination thereof:

    Method paper: An advancement of or extension to an existing computer vision algorithm that has been tailored for animals, showing clear improvements on existing/strong baselines (e.g., the current state-of-the-art algorithm applied to humans if there is no or only little comparable methods for animals yet)
    System paper: A clever combination of existing methods that solves a problem of interest to CV4Animals, rigorously evaluated on two or more datasets and showing that the choice of model components is non-trivial, each component is essential, and that the proposed method compares favorable to the current state-of-the-art.
    Dataset paper: A new dataset that brings about a new aspect (e.g., new species, annotation type, larger scale), comes with a well-define evaluation protocol (train/val/test split and metrics), and provides scores for strong baselines (state-of-the-art on a related task/domain, typically more than one). 
    Other forms are of course possible too, e.g., a state-of-the-art report. The examples just serve as a reference for the most common article types.

This special issue is tied to the United Nation's Sustainable Development Goal 15: Life on Land, aiming to protect, restore, and promote sustainable use of ecosystems and combat animal species extinction. For more information, see: https://sdgs.un.org/goals/goal15 

Timeline

    Final submission deadline: 20 September 2025
    First review decision: 15 December 2025
    Revised paper due: 1 February 2026
    Final review decision: 1 April 2026
    Final manuscript decision: 1 May 2026

Special Issue on Ensuring Trustworthiness in Open-World Visual Recognition
提出日: 2025-12-15

Guest editors

    Hong Liu, The University of Osaka, Japan
    Zhun Zhong, Heifei University of Technology, China
    Wei Ji, Nanjing University, China
    Zhe Zeng, University of Virginia, USA
    Nicu Sebe, University of Trento, Italy
    Tat-Seng Chua, National University of Singapore, Singapore
    Walter Scheirer, University of Notre Dame, USA
    Rita Cucchiara, University of Modena and Reggio Emilia, Italy
    Ming-Hsuan Yang, University of California at Merced, USA

Introduction

Visual recognition is a fundamental task in computer vision, with applications ranging from autonomous driving to healthcare. However, conventional visual recognition methods are often limited to recognizing a fixed set of classes, which can lead to significant challenges in real-world applications. Therefore, open-world visual recognition has has emerged as a promising solution to address these challenges. However, although recent efforts have focused on improving the robustness of visual recognition systems in open-world setting, they still struggle with several key trustworthiness issues, which include but are not limited to:

    Unrobustness with Dynamic Environments: Existing visual recognition methods are often vulnerable to dynamic environments, such as varying lighting conditions, occlusions, and weather changes. These challenges can lead to significant performance degradation in real-world applica­tions, particularly in safety-critical domains like autonomous driving.
    Lack of Interpretability: Deep models (such as CNNs and transformers) are often considered black boxes, making it difficult to understand their decision-making processes. This lack of inter­pretability can hinder trust in the system, especially in high-stakes applications like healthcare and security.
    Ethics and Fairness Bias: If there is social bias (such as sample imbalance in gender, race, and region) in the training data, the model may solidify the bias into the recognition results. For example, the recognition accuracy of face recognition systems for people with darker skin tones is significantly lower than that for people with lighter skin tones (a previous study showed that the error gap is more than 10%).
    Privacy and Security Risks: In scenarios involving faces, medical images, biological features, etc., if the original data is not properly protected, there is a risk of leakage or abuse. For example, if facial data from security cameras is illegally obtained, it may lead to identity theft. On the other hand, some recent studies have shown that attackers can infer the characteristics of training data from model outputs and even reconstruct the original images (such as the abuse of generative adversarial networks), threatening data privacy.

    Reliability and Consistency: The performance of existing visual recognition methods can be inconsistent across different domains or scenarios. For example, a model trained on synthetic data may not perform well on real-world data. This inconsistency can lead to unreliable results in ap­plications where accuracy is critical, such as autonomous driving and medical diagnosis. Moreover, when fusing sensor data (e.g., images, radar, and infrared), issues like time asynchrony, spatial calibration errors, or data conflicts may result in decision-making errors.

    Insufficient Dynamic Adaptability: Conventional visual models assume that test data comes from a known set of categories. In the open world, new objects (such as new car models and new species) constantly appear. The model lacks the ability to “recognize the unknown” and may misclassify them as known categories. On the other hand, when the computing power of edge devices (such as drones and mobile phones) is limited, model compression (such as quantization and pruning) may lead to a decrease in robustness and an increase in the misjudgment rate in low-power mode.

Aims & Scope 

This special issue invites innovative research papers that aim to address these challenges and propose novel techniques for ensuring trustworthiness in open-world visual recognition. Potential topics of interest include, but are not limited to:

    Robustness enhancement. This includes methods for improving the robustness of visual recognition systems against adversarial attacks, catastrophic forgetting, domain shifts, and other challenges. 
    Interpretability and explainability. This includes techniques for making visual recognition systems more interpretable and explainable, enabling users to understand the decision-making process of these systems.
    Ethical considerations and fairness. This includes methods for ensuring that visual recognition systems are fair and unbiased, addressing issues related to social bias in training/test data.
    Privacy-preserving techniques. This includes methods for protecting sensitive data in visual recog­nition systems, ensuring that user privacy is maintained.
    Dynamic adaptability. This includes techniques for enabling visual recognition systems to adapt to new classes and domains in real-time, ensuring that they remain effective in dynamic environments.
    Evaluation criteria and benchmarks. This includes the development of new evaluation criteria and benchmarks for assessing the performance of visual recognition systems in open-world scenarios, ensuring that they are robust, interpretable, and fair.

Moreover, this special issue also welcomes papers that focus on:

    Developing new techniques for continual learning, federated learning and multi-modality learning in the context of open-world visual recognition
    Exploring the use of large vision-language models or vision foundation models for open-world visual recognition tasks
    Encourage interdisciplinary integration by incorporating perspectives from fields such as ethics, law, and sociology.
    New applications of relational generative models (such as diffusion models) in data enhancement and privacy protection. 

We encourage submissions that cover a broad range of visual recognition tasks, including but not limited to image classification, object detection, semantic segmentation, action recognition and pose estimation. This special issue will provide a platform for researchers to share their latest findings and contribute to the advancement of trustworthy visual recognition. The contributions in this special is­sue could significantly benefit society by enabling more robust and reliable visual recognition systems, enhancing public safety and security, social impact, and increasing the efficiency of industrial and com­mercial applications. 

Special issue timeline

    Submission deadline: December 15, 2025
    First review notification: February 15, 2026
    Revised submission deadline: April 15, 2026
    Final review notification: May 15, 2026
    Final manuscript due: June 15, 2026
    Publication date: Summer 2026