Microsoft has released a free artificial intelligence tool for images and videos that works with other open-source machine learning and AI platforms. You supply images or video for a dataset and it sorts them out according to how you label them and learns while you are labeling those images or videos.
From the lobe.ai website:
Machine learning made easy
Lobe has everything you need to bring your machine learning ideas to life. Just show it examples of what you want it to learn, and it automatically trains a custom machine learning model that can be shipped in your app.
While this may not make any sense to you right now, particularly if you are on the outside looking in on this whole AI thing, it is a shockingly simple program to use and develop an image or video-based machine learning model.
And here is the best part. No coding. No programming. Mainly drag and drop images or videos, label them, training is already happening in the background. Then correct wrong guesses to retrain on images it missed.
So... what’s this mean to us... if we aren’t directly involved in AI? Simple... if nothing else it will help you grasp what machine learning is about on a basic level. On the practical end... it can help you with things like your photo morgue if you are a photographer as it can learn to differentiate between male and female, long hair, short hair, and things of that sort for a sort and search app.
For my first use, I chose female or male type of comparison from a folder of faces I keep for reference. This gave me a ready supply of data without having to seek it out and it could eventually be of practical use, but I didn’t have that many images.
They suggest 100 to 1000. I put in 14 male and female faces because they were clear, clean pictures. It’s not a huge dataset or reliable dataset by any means but it works for a quick and dirty test. Due to the small amount of data, it was through training as almost as soon as I was through labeling them male and female.
It had one miss. A long-haired male was mistaken for a female. Not bad at all for a first pass. Glad I had the long-haired man in the image set otherwise I would have thought I was 100% right without realizing how easy my model was to fool.
I corrected the image it labeled wrong and just as quickly it was through learning. A subsequent test of dragging in other images showed the improvement already in distinguishing between long and short hair. In writing this I realized I failed to add any short-haired females to the dataset so you can bet there will be errors with that type of picture until retrained.
All in all, it was a great experience for a free download. I learned quite a bit more than I thought I would as I was confused about what it did till I looked at their example page which made me think of that folder full of faces I could use.
While this is not a cutting-edge example of life-changing technology it was a great learning exercise for those of us not in the thick of AI but wanting to learn more about it. Strangely enough... I couldn’t get the Lobe website to function with Chrome, getting a backscreen. I resorted to using the Chromium-based Edge to download the app and peruse the website.
If you want a free and easy way to see what artificial intelligence-based machine learning is about then hop on over to the lobe.ai website, download the free app, and dive in. Head right for the deepwater... because it’s only jacuzzi size and lots of fun.
Lobe AI Website: https://lobe.ai/
M.D. McCallum, aka WarLord, is an international award-winning commercial graphics artist, 3D animator, published author, project director, and webmaster with a freelance career that spans over 20 years. Now retired, M.D. is currently working part-time on writing and select character development projects. You can learn more about MD on his website.