I couldn;t get any pretrain card models to detect the cards in Balatro :/
So I've had to export the cards from the game, then chop the image files into slices so I have each individual card...
Then I trained a model to detect cards with that data..
It thinks the bounding boxes for detected objects are the whole screen..
Chatgpt tells me because I did that the model learned that each card takes up the whole image always.....
Chatgpt wrote me a script to put multiple cards in the same image with the correct label data... so I'm training that now :/
I thought balatro would be easiest game to write an ai for out of all the ones I have installed :/
just getting it to read the screen elements is crazy hard..
I extracted the custom fonts from the game too... guess I need to train a model for that as well....
Need to organise all the photos, ate soo much food.
Tokyo and hakone where my favorites but we didn't see much of hakone just one little loop out spent most of the time relaxing in the thermal bath.
Kyoto was beautiful but ruined a bit by the sheer number of tourists and we stayed in a lovley little traditional house one road over from the Hōkan-ji Temple (Yasaka Pagoda) so we did a fair bit of the area at night when no one was around.
Emptied the airport vending machine of monster cubed as couldn't find a shop selling and have brought back a bunch of spices and miso and broth base and weird snacks.
Need to plan another trip now and see more of the country side.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.