Visual Bank Inc., located in Shibuya, Tokyo, is excited to announce the launch of its new Qlean Dataset, a comprehensive facial image dataset aimed at facilitating artificial intelligence (AI) development and research. With this initiative, Visual Bank, through its subsidiary Amana Images, seeks to bolster AI training with a diverse array of data options suitable for various applications.
The Qlean Dataset now includes a collection of approximately 1,500 facial images representing individuals of different ages, genders, and ethnic backgrounds, thus enhancing the versatility of AI training data. This expansion is part of the company’s ongoing effort to strengthen its AI Data Recipe offerings, which include a selection of original, commercially viable datasets.
What is the Qlean Dataset?
The Qlean Dataset provides researchers and developers with original data options that are available for commercial use. It allows users to mix and match data materials according to their specific needs, including flexibility in accuracy and delivery timelines. The dataset consists of both annotated and unannotated data, and adjustments can be made based on individual client requirements.
Collaborations with notable partners like the Chiba Lotte Marines and Toyo Keizai Inc. have helped enhance the kit’s offerings, alongside a robust domestic and international network. By leveraging such partnerships, Visual Bank aims to significantly ease the data collection and preparation process for AI development teams, ultimately accelerating project timelines while ensuring high-quality outputs.
Details of the New Dataset
The latest addition to the Qlean Dataset includes:
1.
Multinational Facial Images Dataset (Various Accessories): This consists of images of 110 diverse models captured in various settings with different lighting conditions and accessories such as masks and glasses. The dataset includes 15,232 images, totaling 12.7 GB.
2.
Japanese Facial Images Dataset (Version 1): A collection of images from 200 Japanese individuals across various ages and settings (e.g., with/without accessories) totaling 22,800 images and 3.99 GB.
3.
Japanese Facial Images Dataset (Version 2): Featuring images of 468 Japanese individuals, this variant captures various situations and settings for more than 67,000 images stored with a size of 57.0 GB.
4.
Japanese Facial Images Dataset (Version 3): Including faces of 762 Japanese individuals, this dataset provides a comprehensive collection in various scenarios, set in JPEG format.
Use Cases of the New Datasets
The diverse facial image datasets offer numerous applications:
- - Enhancing Facial Recognition AI: The dataset incorporates images from multiple ethnicities, which helps in reducing bias in AI models while improving the fairness of facial recognition systems.
- - Identification in Masks: By including many masked images, it supports the development of identification systems suitable for high mask-wearing environments like airports and healthcare facilities.
- - AI Training for AR/VR: The high-quality data will facilitate advancements in face tracking and real-time expression recognition, crucial for virtual reality applications.
- - Gesture and Gaze Recognition: Various angles included will help in developing models for human-computer interaction and driver monitoring systems.
- - Surveillance Systems: Utilizing scenarios close to real conditions, it allows for robust evaluation of person identification AI.
- - Metaverse and Avatar Generation: Variations in age, ethnicity, and accessories foster the creation of more individualized avatars for use in gaming and other virtual experiences.
Commitment to Quality and Support
Visual Bank prides itself on providing ethically sourced datasets. Qlean Dataset ensures compliance with global privacy policies by obtaining consent from all subjects involved. This allows for unrestricted use in research and commercial settings, providing peace of mind to developers and researchers alike.
Academic Support Program: Additionally, Visual Bank has launched a program to support academic research by offering over 500,000 data points across 80 different datasets for free to universities and research institutions. This program aims to address the persistent challenges of acquiring high-quality and rights-cleared training data.
About Visual Bank: Founded with a mission to unlock the potential of all data, Visual Bank is a startup focused on creating a new-generation data infrastructure. Other offerings include AI assistance tools and advanced data solutions through its subsidiary, Amana Images. As a participant in the national research development initiative, GENIAC, Visual Bank continues to pursue efforts to support societal needs through innovative AI solutions.
For more details on Qlean Dataset and participation in the academic program, visit
Qlean Dataset Service Site.