Make a massive, searchable online clothing store quickly with machine learning
You don’t have time to be learning how to make machine learning models, you’re a busy entrepreneur! You need to get to a minimally viable product quickly, and solve some use cases for your customers. Thanks to the democratization of AI and machine learning, you might be able to reach MVP a lot faster than before.
Let’s pretend you’re selling lots of clothes on your website, but you don’t necessarily have time to tag every new image that comes in.
People may submit Instagram photos of them wearing your clothes, or you might want to simply make the cataloguing process a whole lot easier.
This is a perfect use case for machine learning. Here’s an example of how I would solve it.
- Get as many sample images of the clothing items you’re interested in categorizing. For example, I downloaded several hundred photos of jeans, pants, shirts, tshirts, sweatshirts, skirts, dresses, and scarves.
- I used by startup’s object recognition model Tagbox because it only takes about 3 minutes to download and run. Also, I can teach it things it doesn’t already know to make it smarter.
- I then used a little script to iterate over all my photos and teach Tagbox what they were.
- Now comes the fun part. I then went and grabbed a new set of totally random images of clothing. Its important that this new, testing set doesn’t have any of the same photos as the ones used to teach Tagbox. I downloaded this set, ran a script to run each new image through Tagbox and voila:
Tagbox has correctly associated the right keywords with the right products/images. I can now crunch through an unlimited number of images of clothing and extract searchable tags to enhance my product.
The bottom line is that I only spent about an hour building the infrastructure to test this capability. That’s a testament to the democratizing of AI brought about by out-of-the-box machine learning companies like Machine Box.
