Hello Hello, Hoping your day is going well! I'm Gesu India, a Computer Science PhD student at Swansea University, UK. I've been working on assistive technologies for several years, and my current research focuses on improving the teachable-object recognizer applications for people who are blind or low-vision in India. If you're a person with visual impairment in India, above the age of 18, and an Android user, I'd highly appreciate you joining me in this endeavor, and help shaping the future of AI for your community, both in India and worldwide.
This project is all about teachable-object recognizers, which are like supercharged versions of object-recognition apps you might have heard of, like Lookout, Seeing AI, or Envision AI. A generic object-recognizer helps you in recognizing generic things in your surroundings, like a chair, table, salt box, shoes, etc. A teachable-object recognizer is an object recognizer that you can customize to your needs by teaching it to recognize your personal things, such as, for finding your personal coffee mug kept somewhere on a crowded table, or your earbuds that have fallen off the table, and more. Even cooler, you can train it to find things that a generic object-recognizer cannot yet, such as objects that are specific to Indian culture, like kitchen utensils or traditional clothing like a saree or bangles. With teachable-object recognizers, you're in charge and have more control over the tool. Now, like generic object-recognizers, teachable-object recognizers may not be very accurate in recognizing things. Why is that the case? Well, because such applications are built using Artificial Intelligence, in short AI, where a computer learns what an object looks like, say a chair or table, using a dataset that contains hundred examples (images or videos) of chairs. Apps like Lookout, Seeing AI, Envision AI are trained using the same dataset that contains hundreds of images for thousands of generic objects. To improve these apps and teachable-object recognizers, we must enrich the dataset with more images for common objects and diversify it. Currently, it's skewed towards objects that are mainly found in the Western countries, so we need to include items from Indian culture like kitchen utensils, sarees, bangles, food items, etc. Importantly, to truly represent the end users, this dataset should be created by individuals who are blind or have low vision. Here's your chance to make a real impact in your community and improve such apps for the blind and low-vision users in India. Join hundreds of people who are blind or have low vision across the globe in this initiative as a ‘data collector’. As a data collector, you will build your own dataset with videos of at least ten objects, such as a wallet, handbag, phone, or earbuds—whatever you want your phone to recognize and find for you when they're misplaced or lost. Your personal dataset will eventually be a part of the global, publicly available dataset for object-recognition technologies, along with the contribution from your community members in India and worldwide. You'll get a two-week window to create your personal dataset using our Android app, "Find My Things", a teachable-object recognizer. During this time, you'll record 8 short (8-10 seconds each) videos per object, for at least ten objects of your choice. The app offers easy-to-follow instructions and audio-haptic feedback for video recording. We estimate that recording videos for these 10 objects should take about 1.5-2 hours, but, just a heads-up, we might ask you to redo a few videos if they don't quite meet the dataset’s quality and privacy standards during the review. We expect that through this process, you will also gain more exposure to AI systems and datasets. For your successful contribution, you can choose between these two: A) Get INR 500 via UPI transfer, or B) Enjoy exclusive access to the 'Find' feature in Find My Things for six months, teaching it more stuff, with no further data being collected during this period. Come join our 'data collector' team and influence the future of AI for your community! We're looking for people who are above 18 years old, legally blind or visually impaired, have no other disabilities affecting participation, have internet access, and have been regular Android smartphone users (with TalkBack) for at least six months. Please note, you can use our app in Hindi or English, as per your preference. If you're interested or have any questions, please reply (only to me) to this email, or email me at 2132...@swansea.ac.uk Don't forget to spread the word with your friends and peers in your community. The more 'data collectors' we have, the bigger and more diverse our dataset becomes. Thanks for your attention so far, and I'm excitedly looking forward to working with you! Kind regards, Gesu India (she/her) PhD student | Computer Science, Swansea University, UK -- Disclaimer: 1. Contents of the mails, factual, or otherwise, reflect the thinking of the person sending the mail and AI in no way relates itself to its veracity; 2. AI cannot be held liable for any commission/omission based on the mails sent through this mailing list.. Search for old postings at: http://www.mail-archive.com/accessindia@accessindia.org.in/ --- You received this message because you are subscribed to the Google Groups "AccessIndia" group. To unsubscribe from this group and stop receiving emails from it, send an email to accessindia+unsubscr...@accessindia.org.in. To view this discussion on the web visit https://groups.google.com/a/accessindia.org.in/d/msgid/accessindia/CAN-VCskJ53hJxpoMnWoS2EVuVAtu1GDU7ywfgWzYD2%2BevHzErg%40mail.gmail.com.
