Launched Caticature - want to know what cat you look like?
Hey Indie Hackers,
Ever wondered what you would look like if you were a cat, or have you ever wanted a cat twin? Caticature, which I just built, takes your uploaded photo and discovers a cat photo that looks like you. (I used DeepAI, NextJS, and The Cat API to put it together.)
As for Caticature itself, you can choose to:
- Dig it 👍
- Find your cat twin 🐱
- Become a magician by turning people into cats 🎱
- Have fun seeing how your friends really look like 😂
A really nice idea just not quiet what I would have expected but more later on this.
My first question is more it feels like the processing takes an extremely long time I've uploaded as test 1k^2 image but it feels like it is taking at least 20-30 seconds to come up with an answer. 1k is not really big so I don't think it is uploading the picture that takes so long.
Overall I think for the average users that is simply too long. Have you benchmarked it and know what is the most time intensive part? Or is it simply the infrastructure it runs on that has not enough power.
Second what I really would have expected is a kind of Image2Image-Transformer so it really transforms your photo portrait into a cat portrait. If you need more detail or help on this feel free to drop me a line I am happy to do so.
You're right, that's a significant flaw in the application. The upload is fast; the slow part is the "find the cat photo" logic.
I originally made this as a sort of small favor for a friend, so I didn't really put too much time into it. I went the lazy way and did a "compare image similarity" function on the uploaded image and an array of cat pictures. It's a hit-or-miss kind of thing. But at least he's somewhat satisfied with it. xD
I have been thinking about looking for ways to improve it, though. The transformer idea would perhaps be, like, finding the human face(s) in the image and morphing it into a cat face? I'm not sure where to start with that, though, so any tips would be appreciated! :)
I see is because it is actively crawling for images or are they from a fixed set? Or is there any way to speed it up? Well what you certainly could do if that set is fixed maybe you can store some of the features you are using to compare the image similarity that would save a lot of time probably.
Well yes actually finding the face would certainly be a first step and then probably crop the photo to that. That is rather trivial to do. You can get face detector easily. Even I've got one (sorry for the self-promo). The other bit I would do not through a morph though that is like a 1990s approach.
Nice project Catherine :) Just one suggestion: I would add the most striking before-after photo above-the-fold (so users can see more easily what they can expect, atm I have to scroll to see the before-and-after photos). I also think you have a pretty good potential for viral marketing here (after a user uploads his cat, to send his friend a request to do the same so they can compare which cat is pretter, etc. :D)
Thanks for taking a look ;) That's a great suggestion! I'll have to figure out how to make the examples more visible to users.