I played with this sample problem this summer. The thing about that problem is that I can generate an arbitrary amount of training data, but still I couldn't make it work. Possibly because I used too small networks or something, but I'm not sure... the current problem is similar, only the card generally takes up a larger area of the image generally.
(The answers suggesting Yolo etc. are not relevant because Yolo doesn't find the center of the card, it just finds the center of the bounding box.)
@GalAster I should try that. It's similar though to the networks that I've tried, only that I put a couple of fully connected layers at the end.
One possibility may be to run Yolo, extract the bounding box, and then run a key point detection network on the extracted image. In that scenario, all cards will more or less take up the entire image. That's more similar to facial key point detection.
I have a list, and I want to go through and partition them into subgroups based on applying a function to the list, by applying the function to increasingly large ranges until the value reaches a cutoff. I.e. if I have list={1,2,2,1,3,1,1,0,1,2,1} and I want to split it into subgroups whose Total is <=3, f[list, Total[#]<=3] = {{1,2},{2,1},{3},{1,1,0,1},{2,1}}. I'm guessing there is a function f for that, but don't know what it is.
Getting all the data in one request is often much faster than sending several hundred requests, because behind the scenes they are made as one request to the server instead of many.
I am getting some message like this in Linux (Ubuntu 18.04) `Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/home/pratip/.local/lib/python3.6/site-packages/wolframclient/evaluation/kernel/kernelsession.py", line 143, in __init__ 'Kernel not found at %s.' % kernel) ` when I run the start a session in python 3.6 `session = WolframLanguageSession('/usr/local/Wolfram/Mathematica/11.3')` @ArnoudBuzing