Because I have different poker cards, I saw the video you have a training system. It's possible public or not?

So I have a homework that has to detect something with openCV. And I'm interested in your study.
I've followed the instructions that you made to detect cards. After messing around with errors. I successfully detected cards. But there's a problem for me:
Not all the time I show the card it can be detected. In your video, you need a dark background. But I'm using my laptop's cam so I just show the cards in front of it. When the lights on my room's off, the program's working just fine. But when I turn on the lights (maybe my background's lighter?), it cannot detect contours and identify cards.
What should I do to make it work all the time?
Thank you for reading. Hope you can help me out!

Hello, it's possible to provide a C# version using EmguCV ?
because i found hard to convert some Python libs like 'numpy' in C#, but 'cv2' it's easy.

[
OpenCV Error: Sizes of input arguments do not match (The operation is neither 'array op array' (where arrays have the same size and the same number of channels), nor 'array op scalar', nor 'scalar op array') in cv::arithm_op, file C:\projects\opencv-python\opencv\modules\core\src\arithm.cpp, line 659
Traceback (most recent call last):
File "C:/Users/Haziq/PycharmProjects/CardDetector/CardDetector.py", line 86, in '<'module'>'
k].suit_diff = Cards.match_card(cards[k], train_ranks, train_suits)
File "D:\OpenCV-Playing-Card-Detector-master\Cards.py", line 260, in match_card
diff_img = cv2.absdiff(qCard.rank_img, Trank.img)
cv2.error: C:\projects\opencv-python\opencv\modules\core\src\arithm.cpp:659: error: (-209) The operation is neither 'array op array' (where arrays have the same size and the same number of channels), nor 'array op scalar', nor 'scalar op array' in function cv::arithm_op
]
I got the error above when a card enter the camera's sight and immediately closed.
p/s= Remove the " ' " from the module.
Thanks in advance.

HI,
Is it possible to make it read the card value based on the edge only(Not the whole card image).

Please give me some guidance on this.
Thanks

This project is so cool.
I tried to run the project on my Mac, it gives error in findContours and LINE_AA syntax. I delete the "dummy" and change LINE_AA to CV_AA, the program finally works. But when it recognize any card, it will putText "Unknow of Unknow" on the card. I am learning C++ not Python, so I don't know how this happens. Can anyone help?

Hi.
I love your thing
I'm just learning python, i tried to set an mss stream as input.
I couldnt get the Rank_suit_detector properly, so i make it run with a folder of png file, that gave me a correct Cards-img folder...
But When i run CardDetector, i get only Unknown of Unknown, and i have to zoom in the card to have some response (but totally inacurate)...
Do you know what i'm struggling with ...?
here the modified rankSuitIsolator

### Takes a card picture and creates a top-down 200x300 flattened image
### of it. Isolates the suit and rank and saves the isolated images.
### Runs through A - K ranks and then the 4 suits.

# Import necessary packages
import cv2
import numpy as np
import time
import Cards
import os
``

from mss import mss
from PIL import Image
mon = {'top': 160, 'left': 160, 'width': 600, 'height': 600}
sct = mss()

imagepath=[
"image/S01.png","image/S01.png","image/S03.png","image/S04.png","image/S05.png","image/S06.png","image/S07.png","image/S08.png","image/S09.png","image/S10.png","image/S11.png","image/S12.png","image/S13.png",
"image/S01.png","image/D05.png",
"image/C04.png","image/H01.png"]
img_path = os.path.dirname(os.path.abspath(__file__)) + '/Card_Imgs/'
IM_WIDTH = 1280;IM_HEIGHT = 720;RANK_WIDTH = 70;RANK_HEIGHT = 125;SUIT_WIDTH = 70;SUIT_HEIGHT = 100

# If using a USB Camera instead of a PiCamera, change PiOrUSB to 2
PiOrUSB = 2

# Use counter variable to switch from isolating Rank to isolating Suit
i = 1

for image in imagepath:
    filename = image
    image = Image.open(image)
    image = np.array(image)
    # print(image)
    # Pre-process image
    gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
    blur = cv2.GaussianBlur(gray,(5,5),0)
    retval, thresh = cv2.threshold(blur,100,255,cv2.THRESH_BINARY)

    # Find contours and sort them by size
    dummy,cnts,hier = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
    cnts = sorted(cnts, key=cv2.contourArea,reverse=True)

    # Assume largest contour is the card. If there are no contours, print an error
    flag = 0
    image2 = image.copy()

    if len(cnts) == 0:
        print('No contours found!')
        quit()

    card = cnts[0]

    # Approximate the corner points of the card
    peri = cv2.arcLength(card,True)
    approx = cv2.approxPolyDP(card,0.01*peri,True)
    pts = np.float32(approx)

    x,y,w,h = cv2.boundingRect(card)

    # Flatten the card and convert it to 200x300
    warp = Cards.flattener(image,pts,w,h)

    # Grab corner of card image, zoom, and threshold
    corner = warp[0:84, 0:32]
    # corner_gray = cv2.cvtColor(corner,cv2.COLOR_BGR2GRAY)
    corner_zoom = cv2.resize(corner, (0,0), fx=4, fy=4)
    corner_blur = cv2.GaussianBlur(corner_zoom,(5,5),0)
    retval, corner_thresh = cv2.threshold(corner_blur, 155, 255, cv2. THRESH_BINARY_INV)

    # Isolate suit or rank
    if i <= 13: # Isolate rank
        rank = corner_thresh[20:185, 0:128] # Grabs portion of image that shows rank
        dummy, rank_cnts, hier = cv2.findContours(rank, cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
        rank_cnts = sorted(rank_cnts, key=cv2.contourArea,reverse=True)
        x,y,w,h = cv2.boundingRect(rank_cnts[0])
        rank_roi = rank[y:y+h, x:x+w]
        rank_sized = cv2.resize(rank_roi, (RANK_WIDTH, RANK_HEIGHT), 0, 0)
        final_img = rank_sized

    if i > 13: # Isolate suit
        suit = corner_thresh[186:336, 0:128] # Grabs portion of image that shows suit
        dummy, suit_cnts, hier = cv2.findContours(suit, cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
        suit_cnts = sorted(suit_cnts, key=cv2.contourArea,reverse=True)
        x,y,w,h = cv2.boundingRect(suit_cnts[0])
        suit_roi = suit[y:y+h, x:x+w]
        suit_sized = cv2.resize(suit_roi, (SUIT_WIDTH, SUIT_HEIGHT), 0, 0)
        final_img = suit_sized

    cv2.imshow("Image",final_img)

    # Save image
    print('Press "c" to continue.')
    key = cv2.waitKey(0) & 0xFF
    if key == ord('c'):
        cv2.imwrite(img_path+filename,final_img)

    i = i + 1

cv2.destroyAllWindows()

Traceback (most recent call last):
File "E:\OpenCV-Playing-Card-Detector-master\CardDetector.py", line 66, in
cnts_sort, cnt_is_card = Cards.find_cards(pre_proc)
File "E:\OpenCV-Playing-Card-Detector-master\Cards.py", line 135, in find_cards
dummy,cnts,hier = cv2.findContours(thresh_image,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
ValueError: not enough values to unpack (expected 3, got 2)

def preprocess_image(image):
    img_w, img_h = np.shape(image)[:2]
    return thresh

should be change to

def preprocess_image(image): 
    img_h, img_w = np.shape(image)[:2]
    return thresh

np.shape() will actually return height than width

So I'm not so hot with python, but would it be possible to use an "if" statement while the image is being compared to the training images to assign the cards "count value" in a string? then later "print" the sum of that string? I know this is your end goal but i just thought i would ask. And thanks this is an awesome project and i'm working on it tonight!

In Windows 10, the default installation of Anaconda 3.6 does not include OpenCV & Numpy libraries. These are required to be downloaded and installed separately from the Anaconda repository.

OpenCV
https://anaconda.org/anaconda/opencv
To install, run "Prompt" and type: conda install -c anaconda opencv

Install Numpy (registration required)
https://anaconda.org/anaconda/numpy
To install, run "Prompt" and type: conda install -c anaconda numpy

Hi there,

I wanted to try this on a screenshot, being succesfull just to decission of cards, my size value of my contours has about 355 and the minimum size is above 25000 I guess.

What do I need to readjust here? have i missed some scaling?

for i in range(len(cnts_sort)):
size = cv2.contourArea(cnts_sort[i]) // has size of 300 and min size is 25000
peri = cv2.arcLength(cnts_sort[i],True)
approx = cv2.approxPolyDP(cnts_sort[i],0.01*peri,True)

Hi,
which software licence is this project offered under?