How exactly does this work?
Search is a powerful tool that lets you upload images to our Cloud (either via our User Interface or our API) and then compare their similarity in some way, shape or form. The algorithm analyzes images at the pixel level and takes a variety of items into account, including shapes, lighting and colors.
What types of searches do you support?
A ton of em! But to be more specific, by...
- A tag (either a Custom or General Model concept, like "person")
- An image in your collection (similar image search)
- Any image outside of your collection (reverse image search)
- A tag + An image (combined)
- A group of images in your collection, or
- Custom Metadata that is uploaded with your images (e.g. Brand, Price, Style, etc.)
So you do support Custom Metadata?
We do! It has to be uploaded via the API at the moment but we are working on adding it to the UI soon as well. Currently it can be searched via either the API or the UI.
Can I use your software to search against my own database of images?
Right now we only offer cloud-based support, but we are hoping to have On-Premise functionality some time in the near future.
Can I upload a picture of a really specific person or object to identify him/her/it?
Unfortunately this is outside the scope of what we do, as we can only readily identify people or objects that our system has been trained on. We hope you find what you're looking for though!
How long do searches generally take?
We can search a collection of 1 million images in nearly half a second! So feel free to upload several million images if you've got em.
What is the ideal size for images that I upload to the platform?
The optimal file and pixel sizes are 500kb and 512x512, though it's totally ok if you can't adhere to those bounds. Just note that really large images in the 5+ megabyte range will start to impact performance, so make sure to compress them ahead of time if you can.
Where can I test this?
Our User Interface is located at https://preview.clarifai.com so feel free to check it out!