#Japan #AI Solutions #Tokyo #Qlean Dataset #Visual Bank

Visual Bank Expands AI Learning Solutions with Qlean Dataset

Visual Bank, a Tokyo-based company, has launched a new addition to its Qlean Dataset—a video and image dataset focusing on gymnastics. This offering is an essential tool for researchers and commercial AI developers seeking high-quality training data. The Qlean Dataset, accessible through Visual Bank's subsidiary, Amana Images, aims to ease the developmental burdens of AI projects by providing flexible and adaptable data solutions. This expansion is part of Visual Bank’s initiative to enhance its uniquely constructed lineup of AI data recipes.

Overview of Qlean Dataset

The Qlean Dataset features original data suitable for commercial use. It allows users to flexibly combine data based on their specific needs, such as accuracy, deadlines, and even customization options based on individual requirements. This adaptability ensures that companies can effectively harness the power of AI in their operations while streamlining the data acquisition process.

The recent addition includes video and image recordings of gymnastic performances from two different angles, showcasing both male and female athletes. The dataset features four gymnasts, two male and two female, all in their twenties, emphasizing the authenticity and relatability of the data for training models.

Key Details:

• - Collection Contents: Videos and images demonstrating gymnastic routines, recorded from two angles.
• - Athlete Demographics: Japanese, male and female gymnasts aged in their 20s.
• - Recording Environment: Indoor settings, specifically gymnasiums.
• - Recorded Actions: Each gymnast performs their respective official routines:

- Men: Floor, pommel horse, rings, parallel bars, and horizontal bar.
- Women: Floor, vault, uneven bars, and balance beam.

Data Specifications

The gymnastics video dataset has a capacity of 15.4GB, with 82 files in MP4 format and a total recording length of 1 hour, 5 minutes, and 21 seconds. For more details, a sample can be found here.

On the other hand, the image dataset aggregates a massive 226.37GB worth of data, encompassing 27,516 images in JPEG format. A detailed overview is shared here.

Use Cases of Gymnastics Video and Image Dataset

The new gymnastics dataset opens various possibilities for applications:

1. Posture Estimation and Skeletal Recognition Models: The dataset covers a diverse range of complex gymnastics movements, making it ideal for training algorithms in skeletal detection and posture estimation. Such models could significantly enhance sports science and rehabilitation research.

2. Sports Science and Educational Research: The dataset can be utilized by universities and research institutions for motion analysis and performance evaluations in sports. Incorporating complex gymnastics movements can lead to improved research accuracy and aid in educational environments.

3. Competitive Scoring and AI Judging Support: Official routine footage facilitates training AI for automatic scoring and performance assessment models. These models could contribute towards ensuring fairness in competitions and providing coaching tools for athlete performance feedback.

4. Motion Generation for VR/AR and the Metaverse: The simultaneous collection from two angles makes this dataset suitable for creating realistic motion simulations in VR and AR environments, enhancing the immersive experience for virtual spectators.

5. Robotics and Motion Imitation Learning: The skills required for events like the rings and pommel horse demand balance and coordination. This dataset is invaluable for robotic motion imitation and humanoid action learning, particularly in reproducing complex movements and enhancing autonomous control.

Academy Support Program

In a bid to contribute towards academia, Visual Bank has initiated a program for the free provision of their datasets to universities, research institutions, and non-profit tech development teams. This program includes over 80 categories of data, exceeding 500,000 points, and aims to address the persistent issue of accessibility to high-quality, rights-cleared training data in research settings. More information can be found here.

Visual Bank Overview

Founded with the mission to unlock the potential of all data, Visual Bank aims to maximize AI development capabilities through a next-generation data infrastructure. The company also offers tools for creators, such as THE PEN, which supports comic artists in enhancing their drawings, and the Qlean Dataset service through its wholly-owned subsidiary, Amana Images. Currently, Visual Bank is part of Japan's governmental research program, GENIAC, accelerating its commitment to social implementation.

About the CEO

The CEO, Masanori Nagai, leads Visual Bank from its headquarters at WeWork, Yotsuya Building, Chiyoda, Tokyo. For more on the company and its offerings, visit Visual Bank and Amana Images.