Coming Soon.
Coming Soon.
The proliferation of deepfakes and AI-generated content has led to a surge in media forgeries and misinformation, necessitating robust detection systems. However, current datasets lack diversity across modalities, languages, and real-world scenarios. To address this gap, we present ILLUSION (Integration of Life-Like Unique Synthetic Identities and Objects from Neural Networks), a large-scale, multi-modal deepfake dataset comprising 1.3 million samples spanning audio-visual forgeries, 26 languages, challenging noisy environments, and various manipulation protocols. Generated using 28 state-of-the-art generative techniques, ILLUSION includes faceswaps, audio spoofing, synchronized audio-video manipulations, and synthetic media while ensuring a balanced representation of gender and skin tone for unbiased evaluation. Using Jaccard Index and UpSet plot analysis, we demonstrate ILLUSION’s distinctiveness and minimal overlap with existing datasets, emphasizing its novel generative coverage. We benchmarked image, audio, video, and multi-modal detection models, revealing key challenges such as performance degradation in multilingual and multi-modal contexts, vulnerability to real-world distortions, and limited generalization to zero-day attacks. By bridging synthetic and real-world complexities, ILLUSION provides a challenging yet essential platform for advancing deepfake detection research.
More details at https://www.iab-rubric.org/illusion-database.
This research explores the bias in text-to-image (TTI) models for the Indic languages widely spoken throughout India. It examines and compares the generative performance and cultural aspects of leading TTI models in these languages, contrasting it with their English language capabilities. Employing the proposed IndicTTI benchmark, this research comprehensively evaluates the performance of 30 Indic languages using two open-source diffusion models and two commercial generation APIs. The primary objective of this benchmark is to measure how well these models support Indic languages and identify areas in need of improvement. Considering the linguistic diversity of 30 languages spoken by over 1.4 billion people, this benchmark aims to provide a detailed and insightful analysis of TTI models' effectiveness in the context of Indic linguistic landscapes.
More details at https://iab-rubric.org/resources/other-databases/indictti.
In the CAAXR dataset, a total of 1,749 chest X-Rays from the publicly available BIMCV database have been annotated by six radiologists from Mahajan Labs (Mahajan Imaging, New Delhi, India). All the six radiologists had more than 5 years experience, with two among them having 10 and 15 years of experience of reading chest X-rays. All the X-rays were annotated only once by a single radiologist. X-rays with doubtful findings were discussed, and consensus resolution was obtained by the independent effort of the annotator.
Four evaluation protocols for CAAXR database are introduced. Three out of four protocols are designed for classification, and one for segmentation. All protocols use 5-fold cross-validation. The baseline results for semantic segmentation use three segmentation models- UNet, SegNet, and Mask-RCNN. The classification performance is evaluated using four deep learning models namely - DenseNet121, MobileNetv2, ResNet18, and VGG19. The models are evaluated using only train and validation splits, ensuring a fair estimate of the models’ performance on the testing data.
The protocol CSVs can be downloaded from here.
The annotations for the chest X-rays are provided here.
The trained baseline segmentation models can be downloaded from the link: Baseline models (10.3 GB)
The Leap Signature Dataset (LSD) consists of signatures from 60 unique subjects in the form of 3D signature trajectories. The trajectories are captured in 3D space without any tactile feedback using a Leap Motion device. The data for each subject consists of 12 samples of genuine signatures and 3 samples of impostor signatures. The data for each subject is captured in a single session. The Leap Signatue Dataset was captured along with the face identities of the subjects. However, to protect the identities of the subjects, we have provided the Leap Signature samples only.
The database can be downloaded from the following link:
Leap Signature Dataset (LSD) (5.34 MB) (CRC32: 5CC395A7, MD5: 26407E44EE078AF134EA114659F29204)
- To obtain the password for the compressed file, email the duly filled license agreement to databases@iab-rubric.org with the subject line "License agreement for Leap Signature Dataset (LSD)"
NOTE: The license agreement has to be signed by someone having the legal authority to sign on behalf of the institute, such as the head of the institution or registrar. If a license agreement is signed by someone else, it will not be processed further.
This database is available only for research and educational purpose and not for any commercial use. If you use the database in any publications or reports, you must refer to the following paper: - I. Nigam, R. Singh, and M. Vatsa. Leap Signature Recognition Using HOOF and HOT Features, In IEEE International Conference On Image Processing, 2014.
- A. Chahar, S. Yadav, I. Nigam, M. Vatsa, and R. Singh. A Leap Password Based Verification System, In Proceedings of IEEE International Conference on Biometrics: Theory, Applications and Systems, 2015.
Bag-of-Lies is a multi-modal dataset consisting of video, audio and eye gaze from 35 unique subjects collected using a carefully designed experiment for the task of automated deception detection (binary classification into truth/lie). It has a total of 325 manually annotated recordings consisting of 162 lies and 163 truths. Along with it, EEG (13 channels) data is also available for 22 unique subjects.
For the experiment, each subject was shown 6-10 select images and were asked to describe them deceptively or otherwise based on their choice. Video and Audio were captured using a standard phone camera and microphone, gaze data was collected using GazePoint GP3 Eye tracking system and EEG Data was captured using EPOC+ headset.
The database can be downloaded from the following link:
(Bag-of-Lies Database) (6.14 GB) (CRC32: 8748CBD7, MD5: A18542168F2F178EBECAA292BFC791B3)
- To obtain the password for the compressed file, email the duly filled license agreement to databases@iab-rubric.org with the subject line "License agreement for Bag-of-Lies Database"
NOTE: The license agreement has to be signed by someone having the legal authority to sign on behalf of the institute, such as the head of the institution or registrar. If a license agreement is signed by someone else, it will not be processed further.
This database is available only for research and educational purpose and not for any commercial use. If you use the database in any publications or reports, you must refer to the following paper: - You must not use data from User_12 (as pictures or otherwise) in any publications / derived works using this dataset.
- You must refer to the following papers:
V. Gupta, M. Agarwal, M. Arora, T. Chakraborty, R. Singh, M. Vatsa. Bag-of-Lies: A Multimodal Dataset for Deception Detection, In IEEE Conference on Computer Vision and Pattern Recognition Workshop on Challenges and Opportunities for Privacy and Security, 2019.
The Head-image Soft Biometric Database (HSBD) consists of head images captured using 2 different DSLR cameras. It has head-images for 103 subjects. This database contains no facial informations and include some angular movements. The database also includes the ground truth annotations of the head in the acquired head-image.
The database can be downloaded from the following link:
Head-imageSoft Biometric Database (HSBD) (2.10 GB) (CRC32: 6CD5276C, MD5: C07C618F3026126C8D4409DD331EED3B)
- To obtain the password for the compressed file, email the duly filled license agreement to databases@iab-rubric.org with the subject line "License agreement for Head-imageSoft Biometric Database (HSBD))"
NOTE: The license agreement has to be signed by someone having the legal authority to sign on behalf of the institute, such as the head of the institution or registrar. If a license agreement is signed by someone else, it will not be processed further.
This database is available only for research and educational purpose and not for any commercial use. If you use the database in any publications or reports, you must refer to the following paper: - A. Malhotra, R. Singh, M. Vatsa, and V. M. Patel. Person Authentication using Head Images, In IEEE Winter Conference on Application of Computer Vision(WACV), 2018.
