US-based training data provider vAIsual has launched a new series of biometric video datasets, providing fully consented materials to accelerate the development of machine learning algorithms.
The new dataset consists of 4k resolution videos of full-body biometrics and postures of models that have signed the biometric data release.
Violation of existing privacy and copyright laws is the biggest challenge for the AI industry, according to vAIsual CEO Michael Osterrieder.
“Cleansing has become standard among generative AI companies trying to overcome the challenges of training AI models on scraped data. We are the only company that can offer it,” Osterrieder said in the announcement.
The video has been added to vAIsual’s dataset shop, which previously offered only still images for training facial recognition and other AI algorithms.
“Adding video opens up a lot of options for the AI team,” comments Osterrieder. “First, the video itself can capture thousands of individual frames and extract them into individual images. These still images offer an enormous amount of variation in expression and movement.”
In addition to biometric algorithms, the dataset can also be used to train generative and diagnostic algorithms, the company said in a statement. The Biometric Image Dataset is specifically designed to be added to the engineer’s AI training workflow.
vAIsual also recently signed a deal to balance the demographic composition of the dataset with Asian stock photography.
article topic
Artificial Intelligence | Biometrics | Data Privacy | Datasets | Face Recognition |
