This site started great. Then a bunch of UI changes occurred and it seems from other threads it’s only gonna get more colorful and bloated.
I really think this needs to be bloated. Please focus on features that enhance function rather than trying to fill every bit of white space with something unnecessary.
Previous iterations didn’t even have pack logos but didn’t cause issues with recognition. Keep things simple and it’ll be a great site.
Hi,
I’m not sure if you are referring to the actual site or to the “suggestions” here on the forum. Please note that the forum suggestions are other people giving site feedback, but that doesn’t mean I will implement those ideas.
If you are talking about the actual site, then please give me some examples of things you don’t like and feedback is appreciated.
I think you are misunderstanding what is being done. The changes being made are only color changes; this is because there are programmers who only focus on the frontend and others on functionality. In my case, I don’t know how specialized card detection libraries work, but I know how to make aesthetic changes. The important thing in a community is that we all contribute, dialogue and add so that the page has more features, works better and is more friendly for new users. I speak for myself, but any contribution to a community, whatever it is, is welcome if it is constructive. I invite you to participate, there are many ways to collaborate besides coding.
It seems there are a few people worried about the UI changes, so I can see where you’re coming from. Like any changes it takes some time to get used to it, and if it really does end up just being too cumbersome, we’re always able to scale it back down (and inevitably some people will be upset that it got changed it again).
In terms of card recognition getting worse, this is simply because there’s a lot more cards to check against. The code for that hasn’t changed much as far as I can tell, but if you think about how many more cards of each type there are, it makes sense why the recognition is worse. For example, there were 32 grass type cards in Genetic Apex. So when you submitted a grass card image, there were only 31 fairly likely “wrong” answers. Now there’s 110 grass type cards! Thats a ~3.4x increase. And that doesn’t include any full art cards which are predominantly green. So unfortunately that tool is going to struggle more with recognition as PTCGP expands.
There’s tradeoffs for everything, and we’re open to suggestions, the more specific the better!
Hi:
From what I see, you know quite a bit about this image detection topic and its implementation. As a hobby, I am making a version of the application in Expo and I have been curious about how to train the TensorFlow model. I have become curious, what should I do to retrain it with the new charts, is it very difficult, what should I do to do it? I would like to contribute, even a little bit.
Hey @JaViJeC, I don’t actually know much about the detection 
, I didn’t participate in that implementation at all (you can check the git blame to see who did though) and have only barely glanced at the code. However, these kinds of things generally are based on how “similar” an image is to another one, where similarity can be pixel by pixel, so given a card that is nearly 50% just green in the grass type case, there’s pretty high similarity. I have no experience training models though. Also, so as not to detract from the original post, we can move this to another thread and perhaps others with more expertise can chime in.