123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197 |
- import argparse
- import json
- from detection.utils import *
- from os.path import splitext, basename, join
- def main():
- ap = argparse.ArgumentParser()
- ap.add_argument("-i", "--input", required=True, help="input image")
- ap.add_argument("-s", "--scale", type=float, default=0.1, help="scale images by this factor (default: x0.1)")
- ap.add_argument("-c", "--config", type=str, default="config.json",
- help="name file to load config (default: config.json)")
- ap.add_argument("-o", "--output", type=str, help="give a directory name to save the game files in it")
- ap.add_argument("--hide", action='store_true', default=False, help="hide images")
- args = vars(ap.parse_args())
- with open(args['config'], 'r') as file:
- config = json.load(file)
- orig, blob_mask, blob, food_mask, food_img, food_list = detect(args['input'], config)
- discrete_img, discrete_blob, discrete_food_list, known_food = discretize(blob, food_list, config['Discrete Width'],
- config['Discrete Height'])
- if args['output'] is not None:
- filename = splitext(basename(args['input']))[0]
- dir = args['output']
- save(join(dir, filename), discrete_img, discrete_blob, discrete_food_list, known_food)
- if not args['hide']:
- print_results(orig, blob_mask, blob, food_mask, food_img, discrete_img, args['scale'])
- def detect(input_file, config):
- img = cv2.imread(input_file)
- height, width, _ = img.shape
- aspect_ratio = config["Aspect Ratio"]
- height = int(width*aspect_ratio)
- """ Resize image to limits in config file """
- limits = np.array(config['Limits'])
- transform_mat = cv2.getPerspectiveTransform(np.float32(limits), np.float32(
- [[0, 0], [width, 0], [width, height], [0, height]]))
- img = cv2.warpPerspective(img, transform_mat, (width, height))
- """ Prepare upper and lower mask board """
- upper_mask = np.zeros(img.shape[0:2], np.uint8)
- lower_mask = np.zeros(img.shape[0:2], np.uint8)
- upper_mask = cv2.rectangle(upper_mask, (0, 0), (width, int(height/2)), 255, thickness=cv2.FILLED)
- lower_mask = cv2.rectangle(lower_mask, (0, int(height / 2)), (width, height), 255, thickness=cv2.FILLED)
- """ Find blob """
- sat = saturation(img)
- # sum = mean_image([satA, satB])
- blob_mask = find_blob(sat)
- blob = cv2.bitwise_and(img, img, mask=blob_mask)
- """ Print blob information """
- print("Blob covering:")
- print("\t{:.2f}% of the board.".format(mean_percent_value(blob_mask)))
- print("\t{:.2f}% of the upper board.".format(
- mean_percent_value(cv2.bitwise_and(blob_mask, blob_mask, mask=upper_mask), img_ratio=0.5)))
- print("\t{:.2f}% of the lower board.".format(
- mean_percent_value(cv2.bitwise_and(blob_mask, blob_mask, mask=lower_mask), img_ratio=0.5)))
- """ Find food """
- food_list, food_mask, food_img = find_food(img, 100, config['Low Food Color'], config['High Food Color'])
- """ Print food information """
- print("Total food discovered: " + str(len(food_list)))
- for i, food in enumerate(food_list):
- print("\tFood N°" + str(i) + ": " + str(food))
- return img, blob_mask, blob, food_mask, food_img, food_list
- def print_results(orig, blob_mask, blob, food_mask, food, discrete, scale=1.0):
- padding = 35
- nbr_width = 2
- nbr_height = 3
- font = cv2.FONT_HERSHEY_SIMPLEX
- fontsize = 0.45
- thickness = 1
- scaled_height = int(orig.shape[0]*scale)
- scaled_width = int(orig.shape[1]*scale)
- pad = np.zeros((scaled_height, padding, orig.shape[2]), dtype=np.uint8)
- line_pad = np.zeros((padding, (scaled_width + padding) * nbr_width + padding, orig.shape[2]), dtype=np.uint8)
- print_img = cv2.resize(orig, (scaled_width, scaled_height))
- middle = ((0, int(scaled_height/2)), (scaled_width, int(scaled_height/2)))
- cv2.line(print_img, middle[0], middle[1], (0, 255, 0), thickness=1)
- cv2.putText(print_img, 'Mid Line', (middle[0][0] + 5, middle[0][1] - 5),
- font, fontsize, (0, 255, 0), thickness, cv2.LINE_AA)
- print_blob_mask = cv2.resize(cv2.cvtColor(blob_mask, cv2.COLOR_GRAY2BGR), (scaled_width, scaled_height))
- print_blob = cv2.resize(blob, (scaled_width, scaled_height))
- print_food_mask = cv2.resize(cv2.cvtColor(food_mask, cv2.COLOR_GRAY2BGR), (scaled_width, scaled_height))
- print_food = cv2.resize(food, (scaled_width, scaled_height))
- print_discrete = cv2.resize(discrete, (scaled_width, scaled_height))
- concat_line1 = np.concatenate((pad, print_img, pad, print_discrete, pad), axis=1)
- concat_line2 = np.concatenate((pad, print_blob_mask, pad, print_blob, pad), axis=1)
- concat_line3 = np.concatenate((pad, print_food_mask, pad, print_food, pad), axis=1)
- aggregate = np.concatenate((line_pad, concat_line1, line_pad, concat_line2, line_pad, concat_line3, line_pad))
- cv2.putText(aggregate, 'Original:',
- (0 * (scaled_width + padding) + padding + 5, 0 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.putText(aggregate, 'Discrete:',
- (1 * (scaled_width + padding) + padding + 5, 0 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.putText(aggregate, 'Blob Mask:',
- (0 * (scaled_width + padding) + padding + 5, 1 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.putText(aggregate, 'Blob:',
- (1 * (scaled_width + padding) + padding + 5, 1 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.putText(aggregate, 'Food Mask:',
- (0 * (scaled_width + padding) + padding + 5, 2 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.putText(aggregate, 'Food Regions:',
- (1 * (scaled_width + padding) + padding + 5, 2 * (scaled_height + padding) + padding - 5),
- font, fontsize, (255, 255, 255), thickness, cv2.LINE_AA)
- cv2.imshow("Results", aggregate)
- print("\nPress any key...")
- cv2.waitKey(0)
- def discretize(blob_img, food_list, width, height):
- img_height, img_width, _ = blob_img.shape
- discrete_blob = cv2.resize(blob_img, (width, height), interpolation=cv2.INTER_NEAREST)
- discrete_food_list = []
- for food in food_list:
- x = int((food[0] + food[2]/2) / img_width * width)
- y = int((food[1] + food[3]/2) / img_height * height)
- discrete_food_list.append((x,y))
- height, width, _ = discrete_blob.shape
- discrete_blob = cv2.cvtColor(discrete_blob, cv2.COLOR_BGR2GRAY)
- known_food = []
- for food in discrete_food_list:
- if discrete_blob[food[1], food[0]] != 0:
- known_food.append([food[0], food[1]])
- discrete_blob_bgr = cv2.cvtColor(discrete_blob, cv2.COLOR_GRAY2BGR)
- discrete_img = cv2.resize(discrete_blob_bgr, (0, 0), fx=10, fy=10, interpolation=cv2.INTER_NEAREST)
- for (x, y) in discrete_food_list:
- cv2.rectangle(discrete_img, (x * 10, y * 10), ((x + 1) * 10, (y + 1) * 10), (0, 255, 0), thickness=cv2.FILLED)
- for (x, y) in known_food:
- cv2.drawMarker(discrete_img, (x * 10 + 5, y * 10 + 5), (255, 255, 255), thickness=2, markerSize=9,
- markerType=cv2.MARKER_TILTED_CROSS)
- return discrete_img, discrete_blob, discrete_food_list, known_food
- def save(filename, discrete_img, discrete_blob, discrete_food_list, known_food):
- height, width = discrete_blob.shape
- board_str = str(width) + ' ' + str(height) + '\n'
- for x in range(height):
- for y in range(width):
- board_str += format(discrete_blob[x, y] != 0, 'd') + "," + format((y, x) in discrete_food_list, 'd') \
- + "," + str(discrete_blob[x, y]) + " "
- board_str = board_str[:-1]
- board_str += "\n"
- board_str = board_str[:-1]
- with open(filename + ".board", 'w') as board_file:
- board_file.write(board_str)
- with open(filename + ".blob", 'w') as blob_file:
- knowledge = dict()
- knowledge['food'] = known_food
- knowledge['max_scouters'] = len(known_food)
- json.dump(knowledge, blob_file)
- with open(filename + ".player", 'w') as player_file:
- player_file.write("0")
- cv2.imwrite(filename + ".jpg", discrete_img)
- if __name__ == "__main__":
- main()
|