Enhancing Object Recognition Algorithms Using Modest Datasets
New Dataset Aims to Improve AI for the Visually Impaired
Microsoft and City, University of London have teamed up to release ORBIT, a unique dataset of annotated videos designed to better train future object recognition systems. The dataset, which consists of 3,822 videos of 486 objects, was recorded by individuals who are blind or have low vision, offering a valuable perspective on the real-world challenges faced by this community.
The videos in ORBIT were shot on smartphones in varied environments to replicate real-world conditions. This dataset is not publicly available for general use but is intended for training AI applications. It is important to note that the videos are not intended for public consumption, and their use is restricted to training purposes.
ORBIT is not a specific AI application but a dataset that can be used to train such applications. The dataset is intended to assist visually impaired individuals by improving the accuracy and effectiveness of object recognition systems.
To access the ORBIT dataset, you can start by visiting the official project or lab webpages of Microsoft Research and City, University of London. Look for a dedicated dataset or resources section, often linked in research papers or project announcements. You may also want to check repositories like GitHub, institutional data archives, or platforms like IEEE DataPort or similar.
In some cases, dataset access may require filling out a request form or agreeing to terms of use due to privacy considerations related to videos featuring people. If you're unable to find the dataset after searching the official sites, you can try looking up relevant academic publications or preprints by the authors of the ORBIT dataset announcement. These often contain dataset access information in their supplementary materials.
Alternatively, you can contact the project leads or corresponding authors via email for guidance on dataset access. If you need help drafting such a request or finding the official research paper linked to ORBIT, feel free to ask!
The ORBIT dataset is an exciting development in the field of AI, offering a valuable resource for improving object recognition systems and, in turn, enhancing the lives of visually impaired individuals.
The ORBIT dataset, created through a collaboration between Microsoft and City, University of London, is specifically designed to train artificial-intelligence (AI) applications, using data collected by visually impaired individuals to better understand the real-world challenges they face. This data (ORBIT dataset) aims to improve AI technology, particularly object recognition systems, for the visually impaired.
