lel

main.py

@@ -0,0 +1,538 @@
+from functools import lru_cache
+
+import cv2
+import runner
+from sklearn.externals import joblib
+import numpy as np
+import operator
+import glob
+import os
+import heapq
+import math
+
+
+pieces = ['rook', 'knight']
+#pieces = ['rook', 'knight']
+#piece_to_symbol = {'rook': 1, 'knight': 2, 'empty': 0}
+piece_to_symbol = {'rook': 1, 'knight': 2}
+colors = ['black', 'white']
+
+
+
+def classify(image, sift : cv2.xfeatures2d_SIFT, file, rank, empty_bias=False):
+    centers = np.load("training_data/centers.npy")
+    probs = {'rook': {'black': 0, 'white': 0}, 'knight': {'black': 0, 'white': 0}, 'empty': {'black': 0, 'white': 0}}
+    #probs = {'rook': 0, 'knight': 0, 'empty': 0}
+    for piece in pieces:
+        for color in colors:
+            #color = runner.compute_color(file, rank)
+            classifier = joblib.load(f"classifiers/classifier_{piece}/{color}.pkl")
+            features = runner.generate_bag_of_words(image, centers, sift)
+            prob = classifier.predict_proba(features)
+
+
+
+            probs[piece][color] = prob[0, 1]
+
+
+    if empty_bias:
+        probs['empty'] *= 1.2
+
+    return probs
+
+
+def pred_test(file, rank, mystery_image=None, empty_bias=False):
+    sift = cv2.xfeatures2d.SIFT_create()
+    if mystery_image is None:
+        mystery_image = cv2.imread("training_images/rook/white/rook_training_D4_2.png")
+    probs = classify(mystery_image, sift, file, rank, empty_bias=empty_bias)
+    return probs
+
+
+
+def pre_process_and_train():
+    runner.do_pre_processing()
+    runner.train_pieces_svm()
+
+
+def build_board_from_dict(board_dict : dict):
+    sift = cv2.xfeatures2d.SIFT_create()
+    board = [[0]*8 for _ in range(8)]
+    counter = 0
+    for idx, value in enumerate(board_dict.values()):
+        probs = classify(value, sift)
+        likely_piece = max(probs.items(), key=operator.itemgetter(1))[0]
+        symbol = piece_to_symbol[likely_piece]
+
+        column = idx // 8
+        row = (idx % 7)
+
+
+        board[row][column] = symbol
+
+        print(probs)
+
+        if likely_piece != 'empty':
+            counter += 1
+
+    print(counter)
+    print(64/(counter-1))
+    return board
+
+
+
+
+def detect_using_nn(spec_image):
+    probs = {'rook': 0, 'knight': 0}
+    for piece in pieces:
+        piece_class = piece_to_symbol[piece]
+        win_size = (64, 64)
+        classifier = joblib.load("classifiers/neural_net_" + piece + ".pkl")
+
+        spec_image = cv2.resize(spec_image, (64, 128))
+        features = np.reshape(spec_image, (1, np.product(spec_image.shape)))
+        prob = classifier.predict_proba(features)
+        print(piece)
+        print(prob[0,1])
+
+
+
+
+def test_entire_board():
+    board = cv2.imread("homo_pls_fuck.jpg")
+    warped = runner.warp_board(board)
+
+    board_dict = runner.get_squares(warped)
+    board = build_board_from_dict(board_dict)
+    print(board)
+
+
+def lel_test():
+    # img = cv2.imread('training_images/rook/white/rook_training_D4_2.png')
+
+    counter = 0
+
+    for filename in glob.glob(os.path.join("training_images", "empty", "*", "*.png")):
+
+        img = cv2.imread(filename)
+        gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+        ret = cv2.adaptiveThreshold(gray, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY, 3, 3)
+
+        # binarize the image
+        #ret, bw = cv2.threshold(gray, 100, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)
+
+        # find connected components
+        connectivity = 4
+        nb_components, output, stats, centroids = cv2.connectedComponentsWithStats(ret, connectivity, cv2.CV_32S)
+        sizes = stats[1:, -1]
+        nb_components = nb_components - 1
+        min_size = 250  # threshhold value for objects in scene
+        img2 = np.zeros((img.shape), np.uint8)
+
+
+        for i in range(0, nb_components + 1):
+            # use if sizes[i] >= min_size: to identify your objects
+            color = np.random.randint(255, size=3)
+            # draw the bounding rectangele around each object
+            cv2.rectangle(img2, (stats[i][0], stats[i][1]), (stats[i][0] + stats[i][2], stats[i][1] + stats[i][3]),
+                          (0, 255, 0), 2)
+            img2[output == i + 1] = color
+
+        #print(nb_components+1)
+
+        if nb_components+1 >= 4:
+            counter += 1
+            print(filename)
+            cv2.imshow("lel", img2)
+            cv2.waitKey(0)
+
+
+    print(counter)
+
+
+def selective_search(image, use_fast=False, use_slow=False):
+    # speed-up using multithreads
+    cv2.setUseOptimized(True)
+    cv2.setNumThreads(4)
+
+
+    if type(image) == str:
+        # read image
+        im = cv2.imread(image)
+    else:
+        im = image
+    # resize image
+    #newHeight = 200
+    #newWidth = int(im.shape[1] * 150 / im.shape[0])
+    #im = cv2.resize(im, (newWidth, newHeight))
+
+    #im = cv2.imread(image)
+
+
+    #lel, im = cv2.threshold(im, 128, 255, cv2.THRESH_BINARY)
+
+
+
+    # create Selective Search Segmentation Object using default parameters
+    ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
+
+    # set input image on which we will run segmentation
+    ss.setBaseImage(im)
+
+    # Switch to fast but low recall Selective Search method
+
+    ss.switchToSingleStrategy()
+
+    if (use_fast):
+        ss.switchToSelectiveSearchFast()
+
+    # Switch to high recall but slow Selective Search method
+    elif (use_slow):
+        ss.switchToSelectiveSearchQuality()
+
+    # run selective search segmentation on input image
+    rects = ss.process()
+    #print('Total Number of Region Proposals: {}'.format(len(rects)))
+
+    # number of region proposals to show
+    numShowRects = 150
+    # increment to increase/decrease total number
+    # of reason proposals to be shown
+    increment = 1
+
+    best_proposals = []
+
+    while True:
+        # create a copy of original image
+
+        # itereate over all the region proposals
+        for i, rect in enumerate(rects):
+            imOut = im.copy()
+
+            # draw rectangle for region proposal till numShowRects
+            if (i < numShowRects):
+                x, y, w, h = rect
+                # cv2.rectangle(imOut, (x, y), (x + w, y + h), (0, 255, 0), 1, cv2.LINE_AA)
+
+                # size = (max(w, x) - min(w, x)) * ((max(h, y) - min(h, y)))
+                top_left = (x,y)
+                bottom_left = (x, y+h)
+                top_right = (x+w, y)
+                bottom_right = (x+w, y+h)
+
+                rect_width = bottom_right[0] - bottom_left[0]
+                rect_height = bottom_right[1] - top_right[1]
+
+
+                size = rect_width * rect_height
+                #print(f"({x}, {y}), ({w}, {h})\n Of size: { size }")
+
+                #cv2.rectangle(imOut, (x, y), (x + w, y + h), (0, 255, 0), 1, cv2.LINE_AA)
+                #cv2.imshow("lel", imOut)
+                #cv2.waitKey(0)
+
+                best_proposals.append((rect, size))
+
+                #if size > biggest_size:
+                #    biggest_rect = (x, y, w, h)
+                #    biggest_size = size
+                #    print(f"New biggest: \n({x}, {y}), ({w}, {h})\nOf size: {biggest_size}")
+
+            else:
+                break
+
+        height, width, channels = im.shape
+        center_x = width // 2
+        center_y = (height // 2)+5
+        dists = []
+        #print(f"Amount of best proposals:\n{len(best_proposals)}")
+
+        #print(f"lel: {len(heapq.nlargest(10, best_proposals, key=lambda x: x[1]))}")
+        for i in heapq.nlargest(10, best_proposals, key=lambda x: x[1]):
+            width, height, channels = im.shape
+            #print(width * height)
+            #print(i[1])
+            x, y, w, h = i[0]
+
+
+
+            if i[1] <= (width*height)*0.8 and i[1] > (width*height)*0.25:
+
+
+                imCop = imOut.copy()
+
+                #cv2.rectangle(imCop, (x, y), (x + w, y + h), (0, 255, 0), 2, cv2.LINE_AA)
+                #cv2.imshow("lel", imCop)
+                #cv2.waitKey(0)
+
+
+                #cv2.rectangle(imCop, (x, y), (x + w, y + h), (0, 255, 0), 4, cv2.LINE_AA)
+
+                top_left = (x,y)
+                bottom_left = (x, y+h)
+                top_right = (x+w, y)
+                bottom_right = (x+w, y+h)
+
+                box_center_x = (top_left[0]+bottom_left[0]+top_right[0]+bottom_right[0]) // 4
+                box_center_y = (top_left[1]+bottom_left[1]+top_right[1]+bottom_right[1]) // 4
+
+                #print(f"{box_center_x}, {box_center_y}, {center_x}, {center_y}")
+
+                dist = (center_x - box_center_x) ** 2 + (center_y - box_center_y) ** 2
+                print(dist)
+                dists.append([i, dist])
+
+
+                cv2.drawMarker(imCop, position=(x+w, h+y), color=(255, 0, 0), thickness=3)
+                cv2.drawMarker(imCop, position=(x+w, y), color=(255, 0, 0), thickness=3)
+                cv2.drawMarker(imCop, position=(x, y), color=(255, 0, 0), thickness=3)
+                cv2.drawMarker(imCop, position=(x, y+h), color=(255, 0, 0), thickness=3)
+
+                cv2.drawMarker(imCop, position=(box_center_x, box_center_y), color=(0, 255, 0), thickness=3)
+
+                cv2.drawMarker(imCop, position=(center_x, center_y), color=(0, 0, 255), thickness=3)
+
+
+                #cv2.imshow("lel", imCop)
+                #cv2.waitKey(0)
+
+        #print("-------"*5)
+
+        for pls in dists:
+            imCop = imOut.copy()
+            x, y, w, h = pls[0][0]
+            #print(x,y,w,h)
+            #print(pls[1])
+
+            top_left = (x, y)
+            bottom_left = (x, y + h)
+            top_right = (x + w, y)
+            bottom_right = (x + w, y + h)
+
+            cv2.drawMarker(imCop, position=(x + w, h + y), color=(255, 0, 0), thickness=3)
+            cv2.drawMarker(imCop, position=(x + w, y), color=(255, 0, 0), thickness=3)
+            cv2.drawMarker(imCop, position=(x, y), color=(255, 0, 0), thickness=3)
+            cv2.drawMarker(imCop, position=(x, y + h), color=(255, 0, 0), thickness=3)
+
+            box_center_x = (top_left[0] + bottom_left[0] + top_right[0] + bottom_right[0]) // 4
+            box_center_y = (top_left[1] + bottom_left[1] + top_right[1] + bottom_right[1]) // 4
+
+            cv2.drawMarker(imCop, position=(box_center_x, box_center_y), color=(0, 255, 0), thickness=3)
+
+            cv2.drawMarker(imCop, position=(center_y, center_x), color=(0, 0, 255), thickness=3)
+
+
+            cv2.rectangle(imCop, (x, y), (x + w, y + h), (0, 255, 0), 2, cv2.LINE_AA)
+            #cv2.imshow("lel", imCop)
+            #cv2.waitKey(0)
+
+        imCop = imOut.copy()
+
+
+
+
+        best = heapq.nsmallest(1, dists, key=lambda x: x[1])
+
+        if (len(best) == 0):
+            return ((0, 0), (0, height), (width, 0), (width, height))
+
+        x, y, w, h = best[0][0][0]
+        cv2.rectangle(imCop, (x, y), (x + w, y + h), (0, 255, 0), 4, cv2.LINE_AA)
+
+        top_left = (x, y)
+        bottom_left = (x, y + h)
+        top_right = (x + w, y)
+        bottom_right = (x + w, y + h)
+
+        cv2.drawMarker(imCop, position=(x + w, h + y), color=(255, 0, 0), thickness=3)
+        cv2.drawMarker(imCop, position=(x + w, y), color=(255, 0, 0), thickness=3)
+        cv2.drawMarker(imCop, position=(x, y), color=(255, 0, 0), thickness=3)
+        cv2.drawMarker(imCop, position=(x, y + h), color=(255, 0, 0), thickness=3)
+
+        box_center_x = (top_left[0] + bottom_left[0] + top_right[0] + bottom_right[0]) // 4
+        box_center_y = (top_left[1] + bottom_left[1] + top_right[1] + bottom_right[1]) // 4
+
+        cv2.drawMarker(imCop, position=(box_center_x, box_center_y), color=(0, 255, 0), thickness=3)
+
+        cv2.drawMarker(imCop, position=(center_x, center_y), color=(0, 0, 255), thickness=3)
+
+
+        #cv2.imshow("lel", imCop)
+        #cv2.waitKey(0)
+        return (top_left, bottom_left, top_right, bottom_right)
+
+        # show output
+        cv2.imshow("Output", imOut)
+
+        # record key press
+        k = cv2.waitKey(0) & 0xFF
+
+        # m is pressed
+        if k == 109:
+            # increase total number of rectangles to show by increment
+            numShowRects += increment
+        # l is pressed
+        elif k == 108 and numShowRects > increment:
+            # decrease total number of rectangles to show by increment
+            numShowRects -= increment
+        # q is pressed
+        elif k == 113:
+            break
+    # close image show window
+    cv2.destroyAllWindows()
+
+def predict(square, file, rank):
+    color = runner.compute_color(file, rank)
+    empty_var_classifier = load_classifier(f"classifiers/classifier_empty_var/{color}.pkl")
+
+    magnitude_of_var = np.linalg.norm(cv2.meanStdDev(square)[1])
+
+    prob = empty_var_classifier.predict_proba(np.array(magnitude_of_var).reshape(-1, 1))
+    print(prob[0, 1])
+    if (prob[0, 1]) > 0.5:
+        return 'empty'
+
+    return None
+
+
+
+@lru_cache()
+def load_classifier(filename):
+    return joblib.load(filename)
+
+
+
+
+if __name__ == '__main__':
+
+    board = cv2.imread("whole_boards/board_102_1554110461.608167_.png")
+    warped = runner.warp_board(board)
+
+    files = "ABCDEFGH"
+    ranks = [1,2,3,4,5,6,7,8]
+
+    counter = 0
+
+    for file in files:
+        for rank in ranks:
+            square = runner.get_square(warped, file, rank)
+            if predict(square, file, rank) == 'empty':
+                counter += 1
+
+    print(counter)
+    exit()
+
+
+    square = runner.get_square(warped, "D", 2)
+
+    gray_square = cv2.cvtColor(square, cv2.COLOR_BGR2GRAY)
+    print(cv2.meanStdDev(gray_square)[1])
+    print(cv2.meanStdDev(square)[1])
+    cv2.imshow("square", square)
+    cv2.waitKey(0)
+
+
+
+
+    print(pred_test("C", 2, square))
+
+    sift: cv2.xfeatures2d_SIFT = cv2.xfeatures2d.SIFT_create()
+    gray = cv2.cvtColor(square, cv2.COLOR_BGR2GRAY)
+
+    kp, desc = sift.detectAndCompute(gray, None)
+
+    cv2.drawKeypoints(square, kp, square)
+
+    cv2.imshow("kp", square)
+    cv2.waitKey(0)
+
+    exit()
+
+    board = cv2.imread("whole_boards/board_202_1554154094.001122_.png")
+
+    runner.fetch_empty_fields(board)
+    exit()
+
+
+    warped = runner.warp_board(board)
+
+    counter = 0
+
+    #square = runner.get_square(warped, "A", 3)
+
+    #top_left, bottom_left, top_right, bottom_right = selective_search(square, use_fast=True)
+    #cropped = square[top_left[1]:bottom_right[1], top_left[0]:bottom_right[0]]
+
+
+
+    for file in files:
+        for rank in ranks:
+
+            square = runner.get_square(warped, file, rank)
+
+
+            top_left, bottom_left, top_right, bottom_right = selective_search(square, use_fast=True)
+            cropped = square[top_left[1]:bottom_right[1], top_left[0]:bottom_right[0]]
+
+            rect_width = bottom_right[0] - bottom_left[0]
+            rect_height = bottom_right[1] - top_right[1]
+
+            size = rect_width * rect_height
+
+            square_height, square_width, channels = square.shape
+
+            empty_bias = (size == square_height*square_width)
+            if size == square_height*square_width:
+                print(f"{file}{rank} is likely empty")
+
+            res = pred_test(file, rank, mystery_image=square, empty_bias=empty_bias)
+            print(res)
+            if (max(res.items(), key=operator.itemgetter(1))[0] == 'empty'):
+                counter += 1
+
+
+
+    print(f"Amount of empty fields: {counter}")
+    #print("Non-cropped:\t",pred_test(square))
+    #print("Cropped:\t",pred_test(cropped))
+
+    #cv2.imshow("square", square)
+    #cv2.waitKey(0)
+
+
+    #runner.do_pre_processing()
+    #runner.train()
+
+
+    #img = "warped_square_B5.png"
+    #detect_using_nn(img)
+
+    #selective_search("training_images/empty/colorless/warped_square_A6.png", use_fast=True)
+    #selective_search("warped_square_B5.png", use_fast=True)
+    img = "training_images/rook/white/rook_training_D4_7.png"
+    #img = "training_images/rook/white_square/rook_training_E4_10.png"
+    #img = "training_images/knight/white_square/training_D5_134.png"
+
+    #top_left, bottom_left, top_right, bottom_right = selective_search(img, use_fast=True)
+    #cropped = cv2.imread(img)[top_left[1]:bottom_right[1], top_left[0]:bottom_right[0]]
+
+    #cv2.imshow("output", cropped)
+    #print(pred_test(cropped))
+    #cv2.waitKey(0)
+
+
+
+    #lel_test()
+
+
+    # test_entire_board()
+
+    #board = [[0, 0, 1, 2, 0, 0, 0, 2], [0, 1, 2, 2, 1, 0, 0, 1], [0, 0, 0, 0, 1, 0, 2, 0], [0, 2, 2, 1, 1, 2, 2, 0], [0, 1, 0, 0, 1, 2, 0, 0], [0, 0, 0, 0, 0, 2, 2, 0], [0, 0, 0, 2, 0, 0, 0, 0], [0, 0, 0, 0, 0, 0, 0, 0]]
+    #for i in board:
+    #    print(i)
+
+    #warped = cv2.imread("homo_pls_fuck.jpg")
+    #square = runner.get_square(warped, "D", 4)
+    #print(pred_test(square))
+    #cv2.imshow("lel", square)
+    #cv2.waitKey(0)

opencv_video.py

@@ -0,0 +1,55 @@
+import itertools
+from pathlib import Path
+from threading import Thread
+from time import sleep
+import numpy as np
+import cv2
+import runner
+from datetime import datetime
+import utils
+
+
+#cap = cv2.VideoCapture(0)
+#cap = cv2.VideoCapture("rtsp://10.192.49.108:8080/h264_ulaw.sdp")
+cap = cv2.VideoCapture(0)
+
+
+piece = "knight"
+color = "black"
+
+
+rank = 8
+
+pieces = {
+    'knight': [("E", rank), ("H", rank)],
+    'rook': [("A", rank), ("F", rank)],
+    'bishop': [("C", rank), ("D", rank)],
+    'king': [("G", rank)],
+    'queen': [("B", rank)]
+}
+
+while(True):
+    # Capture frame-by-frame
+    ret, frame = cap.read()
+
+    # Display the resulting frame
+    cv2.imshow('frame', frame)
+
+    if cv2.waitKey(100) & 0xFF == ord('c'):
+        print(f"capturing frame")
+        # cv2.imwrite(f"single_frame_{counter}.png", frame)
+        utils.imwrite(f"whole_boards/boards_for_empty/board_{datetime.utcnow().timestamp()}_.png", frame)
+        warped = runner.warp_board(frame)
+
+        runner.save_empty_fields(warped, skip_rank=rank)
+
+        for piece, positions in pieces.items():
+            for position in positions:
+                square = runner.get_square(warped, position[0], position[1])
+                x, y = position
+                utils.imwrite(f"training_images/{piece}/{runner.compute_color(x, y)}_square/training_{x}{str(y)}_{datetime.utcnow().timestamp()}.png", square)
+
+
+# When everything done, release the capture
+cap.release()
+cv2.destroyAllWindows()

runner.py

@@ -1,249 +1,438 @@
 import cv2
 import numpy as np
+import glob
+import os
+from sklearn import cluster
+from sklearn import metrics
+from sklearn import svm
+from sklearn.externals import joblib
+from sklearn import neural_network
+import heapq
+from datetime import datetime
+import utils
 
-out_height, out_width = 500, 500
-dstPoints = np.array([(out_height, 0), (0, 0),  (0, out_width), (out_height, out_width)])
 
 
+pieces = ["rook", "knight"]
+colors = ['black', 'white']
 
-img = cv2.imread("IMG_2086.jpeg")
-img2 = cv2.imread("new_baseline_board.png")
-img_tmp = img.copy()
 
-gray_tmp = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
-gray_tmp = np.float32(gray_tmp)
 
-'''
-dst = cv2.cornerHarris(gray_tmp, 20, 3, 0.04)
 
-#result is dilated for marking the corners, not important
-dst = cv2.dilate(dst,None)
 
-# Threshold for an optimal value, it may vary depending on the image.
-img_tmp[dst>0.01*dst.max()]=[0,0,255]
 
-cv2.imwrite('fuck.jpg',img_tmp)
+def selective_search(image, use_fast=False, use_slow=False, image_name = None):
+    # speed-up using multithreads
+    cv2.setUseOptimized(True)
+    cv2.setNumThreads(4)
 
+    im = image
 
+    img_out = im.copy()
 
+    # create Selective Search Segmentation Object using default parameters
+    ss = cv2.ximgproc.segmentation.createSelectiveSearchSegmentation()
 
+    # set input image on which we will run segmentation
+    ss.setBaseImage(im)
 
-img = cv2.GaussianBlur(img,(5,5),0)
+    ss.switchToSingleStrategy()
 
-kernel = np.ones((3,3),np.float32)
-kernel[0,1] = 0
-kernel[0,2] = -1
-kernel[1,0] = 3
-kernel[1,1] = 0
-kernel[1,2] = -3
-kernel[2,1] = 0
-kernel[2,2] = -1
-img = cv2.filter2D(img,-1,kernel)
+    if (use_fast):
+        ss.switchToSelectiveSearchFast()
 
-'''
+    elif (use_slow):
+        ss.switchToSelectiveSearchQuality()
 
+    # run selective search segmentation on input image
+    rects = ss.process()
 
-img_tmp_tmp = img.copy()
-gray_2 = cv2.cvtColor(img_tmp_tmp, cv2.COLOR_BGR2GRAY)
-gray_3 = cv2.cvtColor(img2, cv2.COLOR_BGR2GRAY)
+    # number of region proposals to show
+    numShowRects = 150
 
-MAX_FEATURES = 500
-GOOD_MATCH_PERCENT = 0.0005
+    best_proposals = []
 
+    while True:
+        # create a copy of original image
 
-cv2.imwrite('pls_lasse.jpg', img)
+        # itereate over all the region proposals
+        for i, rect in enumerate(rects):
+            imOut = im.copy()
 
-img_tmp = img.copy()
-gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+            # draw rectangle for region proposal till numShowRects
+            if (i < numShowRects):
+                x, y, w, h = rect
 
-sift = cv2.xfeatures2d.SIFT_create()
-#sift = cv2.ORB_create(MAX_FEATURES)
-#sift = cv2.xfeatures2d.SURF_create()
-kp = sift.detect(gray_2, None)
-kp2 = sift.detect(gray_3, None)
+                top_left = (x,y)
+                bottom_left = (x, y+h)
+                top_right = (x+w, y)
+                bottom_right = (x+w, y+h)
 
-kp, des = sift.compute(gray_2, kp)
-kp2, des2 = sift.compute(gray_3, kp2)
+                rect_width = bottom_right[0] - bottom_left[0]
+                rect_height = bottom_right[1] - top_right[1]
 
-cv2.drawKeypoints(img_tmp_tmp, keypoints=kp, outImage=img_tmp_tmp)
-cv2.imwrite('keypoints_img.jpg', img_tmp_tmp)
+                size = rect_width * rect_height
 
-# FLANN parameters
-FLANN_INDEX_KDTREE = 0
-index_params = dict(algorithm = FLANN_INDEX_KDTREE, trees = 8)
-search_params = dict(checks=100)   # or pass empty dictionary
+                best_proposals.append((rect, size))
 
 
-#matcher = cv2.DescriptorMatcher_create(cv2.DESCRIPTOR_MATCHER_BRUTEFORCE_HAMMING)
-#matches = matcher.match(des, des2, None)
+            else:
+                break
 
+        height, width, channels = im.shape
+        center_x = width // 2
+        center_y = (height // 2)+5
+        dists = []
 
-flann = cv2.FlannBasedMatcher(index_params,search_params)
-matches = flann.knnMatch(des, des2, k=2)
+        for i in heapq.nlargest(10, best_proposals, key=lambda x: x[1]):
+            width, height, channels = im.shape
 
-# Need to draw only good matches, so create a mask
-matchesMask = [[0,0] for i in range(len(matches))]
+            x, y, w, h = i[0]
 
+            if i[1] < (width*height)*0.90 and i[1] > (width*height)*0.25:
 
 
+                top_left = (x,y)
+                bottom_left = (x, y+h)
+                top_right = (x+w, y)
+                bottom_right = (x+w, y+h)
 
-good_matches = []
-# ratio test as per Lowe's paper
-for i,(m,n) in enumerate(matches):
-    if m.distance < 0.5*n.distance:
-        matchesMask[i]=[1,0]
-        good_matches.append([m,n])
+                box_center_x = (top_left[0]+bottom_left[0]+top_right[0]+bottom_right[0]) // 4
+                box_center_y = (top_left[1]+bottom_left[1]+top_right[1]+bottom_right[1]) // 4
 
-draw_params = dict(matchColor = (0,255,0),
-                   singlePointColor = (255,0,0),
-                   matchesMask = matchesMask,
-                   flags = 0)
+                dist = (center_x - box_center_x) ** 2 + (center_y - box_center_y) ** 2
+                dists.append([i, dist])
 
-img3 = cv2.drawMatchesKnn(img_tmp_tmp, kp, img2, kp2, matches, None, **draw_params)
-cv2.imwrite("matches.jpg", img3)
 
+        imCop = imOut.copy()
 
-matches.sort(key=lambda x: x[0].distance, reverse=False)
-# Remove poor matches
-numGoodMatches = int(len(matches) * GOOD_MATCH_PERCENT)
-#good_matches = matches[:numGoodMatches]
 
-# Extract location of good matches
-points1 = np.zeros((len(good_matches), 2), dtype=np.float32)
-points2 = np.zeros((len(good_matches), 2), dtype=np.float32)
+        print(image_name)
+        best = heapq.nsmallest(1, dists, key=lambda x: x[1])
+        x, y, w, h = best[0][0][0]
+        cv2.rectangle(imCop, (x, y), (x + w, y + h), (0, 255, 0), 4, cv2.LINE_AA)
 
-for i, (m, n) in enumerate(good_matches):
-    points1[i, :] = kp[m.queryIdx].pt
-    points2[i, :] = kp2[m.trainIdx].pt
+        top_left = (x, y)
+        bottom_right = (x + w, y + h)
 
-print(points1)
-print(len(points2))
 
-h, mask = cv2.findHomography(points1, points2, cv2.RANSAC)
+        cropped = img_out[top_left[1]:bottom_right[1], top_left[0]:bottom_right[0]]
+        return cropped
 
-height, width, channels = img2.shape
-im1Reg = cv2.warpPerspective(img_tmp_tmp, h, (width, height))
+        # show output
+        cv2.imshow("Output", imOut)
 
-cv2.imwrite('homo_pls_fuck.jpg', im1Reg)
+        # record key press
+        k = cv2.waitKey(0) & 0xFF
 
-'''
+        # m is pressed
+        if k == 109:
+            # increase total number of rectangles to show by increment
+            numShowRects += increment
+        # l is pressed
+        elif k == 108 and numShowRects > increment:
+            # decrease total number of rectangles to show by increment
+            numShowRects -= increment
+        # q is pressed
+        elif k == 113:
+            break
+    # close image show window
+    cv2.destroyAllWindows()
 
-# Sort matches by score
-matches.sort(key=lambda x: x[0].distance, reverse=False)
 
-# Remove poor matches
-numGoodMatches = int(len(matches) * GOOD_MATCH_PERCENT)
-matches = matches[:numGoodMatches]
+def generate_centers(number_of_clusters, sift : cv2.xfeatures2d_SIFT):
+    features = None
+    for piece in pieces:
+        for color in colors:
+            for filename in glob.glob(os.path.join("training_images", piece, f"{color}_square", "*.png")):
+                image = cv2.imread(filename)
+                #image = selective_search(image, use_fast=True)
+                gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+                kp, desc = sift.detectAndCompute(gray, None)
 
-# Draw the top matches
-imMatches = cv2.drawMatches(img_tmp_tmp, kp, img2, kp2, matches, None)
-cv2.imwrite("matches.jpg", imMatches)
+                if features is None:
+                    features = np.array(desc)
+                else:
+                    print(f"{piece}, {color}, {filename}")
+                    features = np.vstack((features, desc))
 
-# Extract location of good matches
-points1 = np.zeros((len(matches), 2), dtype=np.float32)
-points2 = np.zeros((len(matches), 2), dtype=np.float32)
+    k_means = cluster.KMeans(number_of_clusters)
+    k_means.fit(features)
+    return k_means.cluster_centers_
 
-for i, match in enumerate(matches):
-    points1[i, :] = kp[match.queryIdx].pt
-    points2[i, :] = kp2[match.trainIdx].pt
 
-print(len(points1))
-print(len(points2))
-'''
+def generate_bag_of_words(image, centers, sift : cv2.xfeatures2d_SIFT):
+    num_centers = centers.shape[0]
+    histogram = np.zeros((1, num_centers))
 
-'''
-h, mask = cv2.findHomography(points1, points2, cv2.RANSAC)
+    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    kp, desc = sift.detectAndCompute(gray_image, None)
 
-height, width, channels = img2.shape
-im1Reg = cv2.warpPerspective(img_tmp_tmp, h, (width, height))
+    if not kp:
+        return histogram
 
-cv2.imwrite('homo_pls_fuck.jpg', im1Reg)
-'''
+    distances = metrics.pairwise.pairwise_distances(desc, centers)
+    best_centers = np.argmin(distances, axis=1)
 
+    for i in best_centers:
+        histogram[0,i] = histogram[0,i] + 1
+    histogram = histogram / np.sum(histogram)
 
-'''
-gray_tmp = gray.copy()
-gray_tmp = np.float32(gray_tmp)
-dst = cv2.cornerHarris(gray_tmp,10,17,0.1)
+    return histogram
 
-#result is dilated for marking the corners, not important
-dst = cv2.dilate(dst,None)
 
-# Threshold for an optimal value, it may vary depending on the image.
-img_tmp[dst>0.07*dst.max()]=[0,0,255]
+def do_pre_processing():
+    sift = cv2.xfeatures2d.SIFT_create()
+    centers = generate_centers(8, sift)
 
-cv2.imwrite('fuck.jpg',img_tmp)
-'''
+    np.save("training_data/centers", centers)
 
+    for piece in pieces:
+        for color in colors:
+            for filename in glob.glob(os.path.join("training_images", piece, f"{color}_square", "*.png")):
+                image = cv2.imread(filename)
+                #image = selective_search(image, image_name=filename, use_fast=True)
+                bow_features = generate_bag_of_words(image, centers, sift)
+                np.save(f"training_data/{piece}/{color}_square/" + os.path.basename(filename), bow_features)
 
-'''
-ret, corners = cv2.findChessboardCorners(gray, (3,3), None)
 
-imgpoints = []
+def load_training_data(spec_piece, color):
+    X = None
+    Y = None
 
-print(ret)
+    for piece in pieces:
+        piece_class = int(spec_piece == piece)
+        for filename in glob.glob(os.path.join("training_data", piece, f"{color}_square", "*.npy")):
+            data = np.load(filename)
+            if X is None:
+                X = np.array(data)
+                Y = np.array([piece_class])
+            else:
+                X = np.vstack((X, data))
+                Y = np.vstack((Y, [piece_class]))
+    return X, Y
 
-if ret == True:
-    corners2 = cv2.cornerSubPix(gray, corners, (11, 11), (-1, -1), criteria)
-    imgpoints.append(corners)
-    # Draw and display the corners
-    cv2.drawChessboardCorners(img, (3,3), corners2, ret)
-    cv2.imwrite('corners_chess.jpg', img)
-'''
 
-'''
-# Detect edges using Canny
-canny_output = cv2.Canny(gray, 140, 160)
+def train_empty_or_piece_var():
+    pieces = ['empty', 'knight', 'rook']
+    for color in colors:
 
-cv2.imwrite('canny_out.jpg', canny_output)
-'''
-'''
-ret, thresholded = cv2.threshold(gray, 80, 255, cv2.THRESH_BINARY)
-cv2.imwrite('threshold_out.jpg', thresholded)
+        X = None
+        Y = None
 
-lines = cv2.HoughLinesP(canny_output, 0.1, np.pi/60, 1, 30, 20)
-for line in lines:
-    print(line)
-    cv2.line(img, (line[0][0], line[0][1]), (line[0][2], line[0][3]), (0, 0, 255), 2, 8)
+        total_weight = 0
+        for piece in pieces:
+            total_weight += len(glob.glob(os.path.join("training_images", f"{piece}", f"{color}_square", "*.png")))
 
-cv2.imwrite('lined_chess.jpg', img)
+        current_weight = len(glob.glob(os.path.join("training_images", 'empty', f"{color}_square", "*.png")))
 
+        for piece in pieces:
+            piece_class = int('empty' == piece)
+            for filename in glob.glob(os.path.join("training_images", piece, f"{color}_square", "*.png")):
+                img = cv2.imread(filename)
 
+                magnitude_of_var = np.linalg.norm(cv2.meanStdDev(img)[1])
 
-_, contours, hierarchy = cv2.findContours(canny_output, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
-'''
+                if X is None:
+                    X = np.array(magnitude_of_var)
+                    Y = np.array([piece_class])
+                else:
+                    X = np.vstack((X, magnitude_of_var))
+                    Y = np.vstack((Y, [piece_class]))
 
-#cv2.drawContours(img, contours, -1, (255, 0, 0), 2)
+        classifier = svm.SVC(class_weight={0: current_weight, 1: total_weight - current_weight}, probability=True)
+        classifier.fit(X, Y)
+        joblib.dump(classifier, f"classifiers/classifier_empty_var/{color}.pkl")
 
 
-'''
-pls_square = []
-prev_max = -1
-for contour in contours:
-    approx = cv2.approxPolyDP(contour, 0.01 * cv2.arcLength(contour, True), True)
-    if len(approx) == 4:
-        point_set = [(x[0,0], x[0,1]) for x in approx]
-        max_x = max([x[0] for x in point_set])
-        min_x = min([x[0] for x in point_set])
 
-        max_y = max([x[1] for x in point_set])
-        min_y = min([x[1] for x in point_set])
 
-        print(((max_x - min_x) * (max_y - min_y)))
+def train_pieces_svm():
+    for piece in pieces:
+        for color in colors:
+            # TODO: Consider removing empty from total_weights, so all classifiers do not consider empty pieces
+            total_weights = len(glob.glob(os.path.join("training_images", "*", f"{color}_square", "*.png")))
+            current_weight = len(glob.glob(os.path.join("training_images", piece, f"{color}_square", "*.png")))
 
-        pls_square.append(approx)
+            print(f"Trainig for piece: {piece}")
+            X, Y = load_training_data(piece, color)
+            classifier = svm.SVC(class_weight={0: current_weight, 1: total_weights - current_weight}, probability=True)
+            classifier.fit(X, Y)
+            joblib.dump(classifier, f"classifiers/classifier_{piece}/{color}.pkl")
 
-        #if ((max_x - min_x) * (max_y - min_y)) > prev_max:
-        #    prev_max = ((max_x - min_x) * (max_y - min_y))
-        #    pls_square = approx
 
 
-print(pls_square)
-#h, mask = cv2.findHomography(pls_square, dstPoints, cv2.RANSAC)
-#height, width, channels = img.shape
-#warped = cv2.warpPerspective(img, h, (out_height, out_width))
+def compute_features(training_image):
+    sift = cv2.xfeatures2d.SIFT_create()
+    gray_training_image = cv2.cvtColor(training_image, cv2.COLOR_BGR2GRAY)
+    kp = sift.detect(gray_training_image)
+    kp, desc = sift.compute(gray_training_image, kp)
+
+    cv2.drawKeypoints(training_image, kp, training_image)
+
+    return training_image
+
+
+def warp_board(camera_image, debug_image=None):
+    #cv2.imwrite('camera_image.png', camera_image)
+
+    baseline = cv2.imread("new_baseline_board.png")
+
+    camera_image_gray = cv2.cvtColor(camera_image, cv2.COLOR_BGR2GRAY)
+    baseline_gray = cv2.cvtColor(baseline, cv2.COLOR_BGR2GRAY)
+
+    sift = cv2.xfeatures2d.SIFT_create()
+    camera_image_keypoints = sift.detect(camera_image_gray, None)
+    baseline_keypoints = sift.detect(baseline_gray, None)
+
+    camera_image_keypoints, des = sift.compute(camera_image_gray, camera_image_keypoints)
+    baseline_keypoints, des2 = sift.compute(baseline_gray, baseline_keypoints)
+
+    if debug_image is not None:
+        cv2.drawKeypoints(camera_image, keypoints=camera_image_keypoints, outImage=debug_image)
+        cv2.imwrite('keypoints_img.jpg', camera_image)
+
+    # FLANN parameters
+    FLANN_INDEX_KDTREE = 0
+    index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=8)
+    search_params = dict(checks=100)   # or pass empty dictionary
+
+    flann = cv2.FlannBasedMatcher(index_params,search_params)
+    matches = flann.knnMatch(des, des2, k=2)
+
+    # Need to draw only good matches, so create a mask
+    matchesMask = [[0,0] for i in range(len(matches))]
+
+    good_matches = []
+
+    # ratio test as per Lowe's paper
+    for i,(m,n) in enumerate(matches):
+        if m.distance < 0.55*n.distance:
+            matchesMask[i]=[1,0]
+            good_matches.append([m,n])
+
+    draw_params = dict(matchColor=(0,255,0),
+                       singlePointColor=(255,0,0),
+                       matchesMask=matchesMask,
+                       flags=0)
+
+    img3 = cv2.drawMatchesKnn(camera_image,
+                              camera_image_keypoints,
+                              baseline,
+                              baseline_keypoints,
+                              matches,
+                              None,
+                              **draw_params)
+    cv2.imwrite("matches.jpg", img3)
+
+    # Extract location of good matches
+    points1 = np.zeros((len(good_matches), 2), dtype=np.float32)
+    points2 = np.zeros((len(good_matches), 2), dtype=np.float32)
+
+    for i, (m, n) in enumerate(good_matches):
+        points1[i, :] = camera_image_keypoints[m.queryIdx].pt
+        points2[i, :] = baseline_keypoints[m.trainIdx].pt
+
+    # print(len(points2))
+
+    h, mask = cv2.findHomography(points1, points2, cv2.RANSAC)
+
+    height, width, channels = baseline.shape
+    im1Reg = cv2.warpPerspective(camera_image, h, (width, height))
+
+    # cv2.imwrite('homo_pls_fuck.jpg', im1Reg)
+    return im1Reg
+
+
+def get_square(warped_board, file, rank):
+    files = "ABCDEFGH"
+    file = files.index(file)
+    rank = 8 - rank
+    width, _, _ = warped_board.shape  # board is square anyway
+
+    side = int(width * 0.04)
+    size = width - 2 * side
+    square_size = size // 8
+
+    padding = 0
+
+    x1 = side + (square_size * file)
+    x2 = x1 + square_size
+    y1 = max(0, side + (square_size * rank) - padding)
+    y2 = min(width, y1 + square_size + padding)
+
+    square = warped_board[y1:y2, x1:x2]
+    return square
+
+
+def get_squares(warped_board):
+    result = {}
+    for file in "ABCDEFGH":
+        for rank in range(1, 9):
+            square =  get_square(warped_board, file, rank)
+            result[f"{file}{rank}"] = square
+            # cv2.imwrite(f"warped_square_{file}{rank}.png", square)
+    return result
+
+def load_data_nn(spec_piece):
+    X = None
+    Y = None
+    for piece in pieces:
+        piece_class = int(spec_piece == piece)
+        for filename in glob.glob(os.path.join("training_images", piece, "*", "*.png")):
+            image = cv2.imread(filename)
+            image = cv2.resize(image, (64, 128))
+            data = np.reshape(image, (1, np.product(image.shape)))
+            if X is None:
+                if piece_class == 1:
+                    for _ in range(10):
+                        X = np.array(data)
+                        Y = np.array([piece_class])
+                else:
+                    X = np.array(data)
+                    Y = np.array([piece_class])
+            else:
+                if piece_class == 1:
+                    for _ in range(10):
+                        X = np.vstack((X, data))
+                        Y = np.vstack((Y, [piece_class]))
+                else:
+                    X = np.vstack((X, data))
+                    Y = np.vstack((Y, [piece_class]))
+    return (X, Y)
+
+def train_nn():
+    for piece in pieces:
+        X, Y = load_data_nn(piece)
+        classifier = neural_network.MLPClassifier(hidden_layer_sizes=64)
+        classifier.fit(X, Y)
+        joblib.dump(classifier, "classifiers/neural_net_" + piece + ".pkl")
+
+
+def letter_to_int(letter):
+    alphabet = list('ABCDEFGH')
+    return alphabet.index(letter) + 1
+
+def compute_color(file, rank):
+    if ((letter_to_int(file)+rank) % 2):
+        return 'white'
+    else:
+        return 'black'
+
+
+def save_empty_fields(warped, skip_rank=None):
+    alpha = "ABCDEFGH"
+    ranks = [1, 2, 3, 4, 5, 6, 7, 8]
+
+    if skip_rank is not None:
+        ranks.remove(skip_rank)
+
+    for file in alpha:
+        for rank in ranks:
+            square = get_square(warped, file, rank)
+            color = compute_color(file, rank)
+
+            utils.imwrite(f"training_images/empty/{color}_square/training_{file}{rank}_{datetime.utcnow().timestamp()}.png", square)
+
 
-cv2.drawContours(img, contours, -1, (255, 0, 0), 2)
 
-cv2.imwrite('contours_chess.jpg', img)
-#cv2.imwrite('homo_img_chess.jpg', warped)'''