From 504ece5b00f192d5c1b343fd06ce1648f9139180 Mon Sep 17 00:00:00 2001
From: Edmond Yoo <hj3yoo@uwaterloo.ca>
Date: Mon, 17 Sep 2018 03:06:19 +0000
Subject: [PATCH] Code cleaning & training new YOLO model
---
opencv_dnn.py | 357 ++++++++++++++++++++++++++++++++++++++++-------------------
1 files changed, 243 insertions(+), 114 deletions(-)
diff --git a/opencv_dnn.py b/opencv_dnn.py
index 25d5848..9acdb5c 100644
--- a/opencv_dnn.py
+++ b/opencv_dnn.py
@@ -1,13 +1,118 @@
import cv2
import numpy as np
+import pandas as pd
+import imagehash as ih
import os
import sys
import math
-from operator import itemgetter
+import random
+import time
+from PIL import Image
+import fetch_data
+import transform_data
+
+card_width = 315
+card_height = 440
-# Disclaimer: majority of the basic framework in this file is modified from the following tutorial:
-# https://www.learnopencv.com/deep-learning-based-object-detection-using-yolov3-with-opencv-python-c/
+def calc_image_hashes(card_pool, save_to=None):
+ card_pool['art_hash'] = np.NaN
+ for ind, card_info in card_pool.iterrows():
+ if ind % 100 == 0:
+ print(ind)
+ img_name = '%s/card_img/png/%s/%s_%s.png' % (transform_data.data_dir, card_info['set'],
+ card_info['collector_number'],
+ fetch_data.get_valid_filename(card_info['name']))
+ card_img = cv2.imread(img_name)
+ if card_img is None:
+ fetch_data.fetch_card_image(card_info,
+ out_dir='%s/card_img/png/%s' % (transform_data.data_dir, card_info['set']))
+ card_img = cv2.imread(img_name)
+ if card_img is None:
+ print('WARNING: card %s is not found!' % img_name)
+ img_art = Image.fromarray(card_img[121:580, 63:685]) # For 745*1040 size card image
+ art_hash = ih.phash(img_art, hash_size=32, highfreq_factor=4)
+ card_pool.at[ind, 'art_hash'] = art_hash
+ img_card = Image.fromarray(card_img)
+ card_hash = ih.phash(img_card, hash_size=32, highfreq_factor=4)
+ card_pool.at[ind, 'card_hash'] = card_hash
+ card_pool = card_pool[['artist', 'border_color', 'collector_number', 'color_identity', 'colors', 'flavor_text',
+ 'image_uris', 'mana_cost', 'legalities', 'name', 'oracle_text', 'rarity', 'type_line',
+ 'set', 'set_name', 'power', 'toughness', 'art_hash', 'card_hash']]
+ if save_to is not None:
+ card_pool.to_pickle(save_to)
+ return card_pool
+
+
+# www.pyimagesearch.com/2014/08/25/4-point-opencv-getperspective-transform-example/
+def order_points(pts):
+ # initialzie a list of coordinates that will be ordered
+ # such that the first entry in the list is the top-left,
+ # the second entry is the top-right, the third is the
+ # bottom-right, and the fourth is the bottom-left
+ rect = np.zeros((4, 2), dtype="float32")
+
+ # the top-left point will have the smallest sum, whereas
+ # the bottom-right point will have the largest sum
+ s = pts.sum(axis=1)
+ rect[0] = pts[np.argmin(s)]
+ rect[2] = pts[np.argmax(s)]
+
+ # now, compute the difference between the points, the
+ # top-right point will have the smallest difference,
+ # whereas the bottom-left will have the largest difference
+ diff = np.diff(pts, axis=1)
+ rect[1] = pts[np.argmin(diff)]
+ rect[3] = pts[np.argmax(diff)]
+
+ # return the ordered coordinates
+ return rect
+
+
+# www.pyimagesearch.com/2014/08/25/4-point-opencv-getperspective-transform-example/
+def four_point_transform(image, pts):
+ # obtain a consistent order of the points and unpack them
+ # individually
+ rect = order_points(pts)
+ (tl, tr, br, bl) = rect
+
+ # compute the width of the new image, which will be the
+ # maximum distance between bottom-right and bottom-left
+ # x-coordiates or the top-right and top-left x-coordinates
+ widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+ widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+ maxWidth = max(int(widthA), int(widthB))
+
+ # compute the height of the new image, which will be the
+ # maximum distance between the top-right and bottom-right
+ # y-coordinates or the top-left and bottom-left y-coordinates
+ heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+ heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+ maxHeight = max(int(heightA), int(heightB))
+
+ # now that we have the dimensions of the new image, construct
+ # the set of destination points to obtain a "birds eye view",
+ # (i.e. top-down view) of the image, again specifying points
+ # in the top-left, top-right, bottom-right, and bottom-left
+ # order
+ dst = np.array([
+ [0, 0],
+ [maxWidth - 1, 0],
+ [maxWidth - 1, maxHeight - 1],
+ [0, maxHeight - 1]], dtype="float32")
+
+ # compute the perspective transform matrix and then apply it
+ mat = cv2.getPerspectiveTransform(rect, dst)
+ warped = cv2.warpPerspective(image, mat, (maxWidth, maxHeight))
+
+ # If the image is horizontally long, rotate it by 90
+ if maxWidth > maxHeight:
+ center = (maxHeight / 2, maxHeight / 2)
+ mat_rot = cv2.getRotationMatrix2D(center, 270, 1.0)
+ warped = cv2.warpAffine(warped, mat_rot, (maxHeight, maxWidth))
+
+ # return the warped image
+ return warped
# Get the names of the output layers
@@ -19,6 +124,7 @@
# Remove the bounding boxes with low confidence using non-maxima suppression
+# https://www.learnopencv.com/deep-learning-based-object-detection-using-yolov3-with-opencv-python-c/
def post_process(frame, outs, thresh_conf, thresh_nms):
frame_height = frame.shape[0]
frame_width = frame.shape[1]
@@ -34,6 +140,7 @@
class_id = np.argmax(scores)
confidence = scores[class_id]
if confidence > thresh_conf:
+ #print(detection[0:3])
center_x = int(detection[0] * frame_width)
center_y = int(detection[1] * frame_height)
width = int(detection[2] * frame_width)
@@ -86,7 +193,7 @@
# Set all brightness values, where the pixels are still saturated to 0.
v[non_sat == 0] = 0
# filter out very bright pixels.
- glare = (v > 240) * 255
+ glare = (v > 200) * 255
# Slightly increase the area for each pixel
glare = cv2.dilate(glare.astype(np.uint8), disk)
@@ -96,76 +203,42 @@
return corrected
-def find_card(img, thresh_val=80, blur_radius=None, dilate_radius=None, min_hyst=80, max_hyst=200, min_line_length=None, max_line_gap=None, debug=False):
- # Default values
- if blur_radius is None:
- blur_radius = math.floor(min(img.shape[:2]) / 100 + 0.5) // 2 * 2 + 1 # Rounded to the nearest odd
- if dilate_radius is None:
- dilate_radius = math.floor(min(img.shape[:2]) / 67 + 0.5)
- if min_line_length is None:
- min_line_length = min(img.shape[:2]) / 3
- if max_line_gap is None:
- max_line_gap = min(img.shape[:2]) / 10
-
- thresh_radius = math.floor(min(img.shape[:2]) / 50 + 0.5) // 2 * 2 + 1 # Rounded to the nearest odd
-
- print(blur_radius, dilate_radius, thresh_radius, min_line_length, max_line_gap)
- '''
- blur_radius = 3
- dilate_radius = 3
- thresh_radius = 3
- min_line_length = 5
- max_line_gap = 5
- '''
-
+def find_card(img, thresh_c=5, kernel_size=(3, 3), size_ratio=0.2):
+ # Typical pre-processing - grayscale, blurring, thresholding
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
- # Median blur better removes background textures than Gaussian blur
- img_blur = cv2.medianBlur(img_gray, blur_radius)
- # Truncate the bright area while detecting the border
- img_thresh = cv2.adaptiveThreshold(img_blur, 128, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, thresh_radius, 5)
- # _, img_thresh = cv2.threshold(img_blur, thresh_val, 255, cv2.THRESH_TRUNC)
+ img_blur = cv2.medianBlur(img_gray, 5)
+ img_thresh = cv2.adaptiveThreshold(img_blur, 255, cv2.ADAPTIVE_THRESH_MEAN_C, cv2.THRESH_BINARY_INV, 5, thresh_c)
- # Dilate the image to emphasize thick borders around the card
- kernel_dilate = np.ones((dilate_radius, dilate_radius), np.uint8)
- img_dilate = cv2.dilate(img_thresh, kernel_dilate, iterations=1)
- img_dilate = cv2.erode(img_dilate, kernel_dilate, iterations=1)
+ # Dilute the image, then erode them to remove minor noises
+ kernel = np.ones(kernel_size, np.uint8)
+ img_dilate = cv2.dilate(img_thresh, kernel, iterations=1)
+ img_erode = cv2.erode(img_dilate, kernel, iterations=1)
- img_contour = img_dilate.copy()
- _, contours, _ = cv2.findContours(img_contour, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
- img_contour = cv2.cvtColor(img_contour, cv2.COLOR_GRAY2BGR)
- img_contour = cv2.drawContours(img_contour, contours, -1, (128, 0, 0), 1)
+ # Find the contour
+ #img_contour = img_erode.copy()
+ _, cnts, hier = cv2.findContours(img_erode, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
+ if len(cnts) == 0:
+ print('no contours')
+ return []
+ #img_contour = cv2.cvtColor(img_contour, cv2.COLOR_GRAY2BGR)
+ #img_contour = cv2.drawContours(img_contour, cnts, -1, (0, 255, 0), 1)
+ #cv2.imshow('test', img_contour)
- # find the biggest area
- c = max(contours, key=cv2.contourArea)
+ # For each contours detected, check if they are large enough and are rectangle
+ cnts_rect = []
+ ind_sort = sorted(range(len(cnts)), key=lambda i: cv2.contourArea(cnts[i]), reverse=True)
+ for i in range(len(cnts)):
+ size = cv2.contourArea(cnts[ind_sort[i]])
+ peri = cv2.arcLength(cnts[ind_sort[i]], True)
+ approx = cv2.approxPolyDP(cnts[ind_sort[i]], 0.04 * peri, True)
+ if size > img.shape[0] * img.shape[1] * size_ratio and len(approx) == 4:
+ cnts_rect.append(approx)
- x, y, w, h = cv2.boundingRect(c)
- # draw the book contour (in green)
- img_contour = cv2.drawContours(img_contour, [c], -1, (0, 255, 0), 1)
-
- # Canny edge - low minimum hysteresis to detect glowed area,
- # and high maximum hysteresis to compensate for high false positives.
- img_canny = cv2.Canny(img_dilate, min_hyst, max_hyst)
-
- detected_lines = cv2.HoughLinesP(img_dilate, 1, np.pi / 180, threshold=300,
- minLineLength=min_line_length,
- maxLineGap=max_line_gap)
- card_found = detected_lines is not None
- if card_found:
- print(len(detected_lines))
-
- img_hough = cv2.cvtColor(img_canny.copy(), cv2.COLOR_GRAY2BGR)
- if card_found:
- for line in detected_lines:
- x1, y1, x2, y2 = line[0]
- cv2.line(img_hough, (x1, y1), (x2, y2), (0, 0, 255), 1)
-
- img_thresh = cv2.cvtColor(img_thresh, cv2.COLOR_GRAY2BGR)
- img_dilate = cv2.cvtColor(img_dilate, cv2.COLOR_GRAY2BGR)
- #img_canny = cv2.cvtColor(img_canny, cv2.COLOR_GRAY2BGR)
- return img_thresh, img_dilate, img_contour, img_hough
+ return cnts_rect
-def detect_frame(net, classes, img, thresh_conf=0.5, thresh_nms=0.4, in_dim=(416, 416), display=True, out_path=None):
+def detect_frame(net, classes, img, thresh_conf=0.1, thresh_nms=0.4, in_dim=(416, 416), out_path=None, display=True,
+ debug=False):
img_copy = img.copy()
# Create a 4D blob from a frame.
blob = cv2.dnn.blobFromImage(img, 1 / 255, in_dim, [0, 0, 0], 1, crop=False)
@@ -176,74 +249,107 @@
# Runs the forward pass to get output of the output layers
outs = net.forward(get_outputs_names(net))
+ img_result = img.copy()
+
# Remove the bounding boxes with low confidence
obj_list = post_process(img, outs, thresh_conf, thresh_nms)
for obj in obj_list:
class_id, confidence, box = obj
left, top, width, height = box
- draw_pred(img, class_id, classes, confidence, left, top, left + width, top + height)
+ draw_pred(img_result, class_id, classes, confidence, left, top, left + width, top + height)
# Put efficiency information. The function getPerfProfile returns the
# overall time for inference(t) and the timings for each of the layers(in layersTimes)
- t, _ = net.getPerfProfile()
- label = 'Inference time: %.2f ms' % (t * 1000.0 / cv2.getTickFrequency())
- cv2.putText(img, label, (0, 15), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255))
+ #if display:
+ # t, _ = net.getPerfProfile()
+ # label = 'Inference time: %.2f ms' % (t * 1000.0 / cv2.getTickFrequency())
+ # cv2.putText(img_result, label, (0, 15), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255))
+
+ '''
+ Assuming that the model has properly identified all cards, there should be 1 card that can be classified per
+ bounding box. Find the largest rectangular contour from the region of interest, and identify the card by
+ comparing the perceptual hashing of the image with the other cards' image from the database.
+ '''
+ card_name_list = []
+ for i in range(len(obj_list)):
+ _, _, box = obj_list[i]
+ left, top, width, height = box
+ # Just in case the bounding box trimmed the edge of the cards, give it a bit of offset around the edge
+ offset_ratio = 0.1
+ x1 = max(0, int(left - offset_ratio * width))
+ x2 = min(img.shape[1], int(left + (1 + offset_ratio) * width))
+ y1 = max(0, int(top - offset_ratio * height))
+ y2 = min(img.shape[0], int(top + (1 + offset_ratio) * height))
+ img_snip = img[y1:y2, x1:x2]
+ cnts = find_card(img_snip)
+ if len(cnts) > 0:
+ cnt = cnts[0] # The largest (rectangular) contour
+ pts = np.float32([p[0] for p in cnt])
+ img_warp = four_point_transform(img_snip, pts)
+ img_warp = cv2.resize(img_warp, (card_width, card_height))
+ '''
+ img_art = img_warp[47:249, 22:294]
+ img_art = Image.fromarray(img_art.astype('uint8'), 'RGB')
+ art_hash = ih.phash(img_art, hash_size=32, highfreq_factor=4)
+ card_pool['hash_diff'] = card_pool['art_hash'] - art_hash
+ min_cards = card_pool[card_pool['hash_diff'] == min(card_pool['hash_diff'])]
+ card_name = min_cards.iloc[0]['name']
+ '''
+ img_card = Image.fromarray(img_warp.astype('uint8'), 'RGB')
+ card_hash = ih.phash(img_card, hash_size=32, highfreq_factor=4)
+ card_pool['hash_diff'] = card_pool['card_hash'] - card_hash
+ min_cards = card_pool[card_pool['hash_diff'] == min(card_pool['hash_diff'])]
+ card_name = min_cards.iloc[0]['name']
+ card_name_list.append(card_name)
+ hash_diff = min_cards.iloc[0]['hash_diff']
+
+ # Display the result
+ if debug:
+ # cv2.rectangle(img_warp, (22, 47), (294, 249), (0, 255, 0), 2)
+ cv2.putText(img_warp, card_name + ', ' + str(hash_diff), (0, 50),
+ cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2)
+ cv2.putText(img_result, card_name , (x1, y1), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 255), 2)
+ if debug:
+ cv2.imshow('card#%d' % i, img_warp)
+ elif debug:
+ cv2.imshow('card#%d' % i, np.zeros((1, 1), dtype=np.uint8))
if out_path is not None:
- cv2.imwrite(out_path, img.astype(np.uint8))
- if display:
- no_glare = remove_glare(img_copy)
- img_concat = np.concatenate((img, no_glare), axis=1)
- cv2.imshow('result', img_concat)
- for i in range(len(obj_list)):
- class_id, confidence, box = obj_list[i]
- left, top, width, height = box
- img_snip = img_copy[max(0, top):min(img.shape[0], top + height),
- max(0, left):min(img.shape[1], left + width)]
- img_thresh, img_dilate, img_canny, img_hough = find_card(img_snip)
- img_concat = np.concatenate((img_snip, img_thresh, img_dilate, img_canny, img_hough), axis=1)
- cv2.imshow('feature#%d' % i, img_concat)
- cv2.waitKey(0)
- cv2.destroyAllWindows()
+ cv2.imwrite(out_path, img_result.astype(np.uint8))
- return obj_list
+ return obj_list, card_name_list, img_result
-def detect_video(net, classes, capture, thresh_conf=0.5, thresh_nms=0.4, in_dim=(416, 416), display=True, out_path=None):
+def detect_video(net, classes, capture, thresh_conf=0.5, thresh_nms=0.4, in_dim=(416, 416), out_path=None, display=True,
+ debug=False):
if out_path is not None:
vid_writer = cv2.VideoWriter(out_path, cv2.VideoWriter_fourcc('M', 'J', 'P', 'G'), 30,
(round(capture.get(cv2.CAP_PROP_FRAME_WIDTH)),
round(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))))
max_num_obj = 0
while True:
+ start_time = time.time()
ret, frame = capture.read()
if not ret:
# End of video
print("End of video. Press any key to exit")
cv2.waitKey(0)
break
- img = frame.copy()
- obj_list = detect_frame(net, classes, frame, thresh_conf=thresh_conf, thresh_nms=thresh_nms, in_dim=in_dim,
- display=False, out_path=None)
- max_num_obj = max(max_num_obj, len(obj_list))
- if display:
- no_glare = remove_glare(img)
- img_concat = np.concatenate((frame, no_glare), axis=1)
- cv2.imshow('result', img_concat)
- for i in range(len(obj_list)):
- class_id, confidence, box = obj_list[i]
- left, top, width, height = box
- img_snip = img[max(0, top):min(img.shape[0], top + height),
- max(0, left):min(img.shape[1], left + width)]
- img_thresh, img_dilate, img_canny, img_hough = find_card(img_snip)
- img_concat = np.concatenate((img_snip, img_thresh, img_dilate, img_canny, img_hough), axis=1)
- cv2.imshow('feature#%d' % i, img_concat)
+ # Use the YOLO model to identify each cards annonymously
+ obj_list, card_name_list, img_result = detect_frame(net, classes, frame, thresh_conf=thresh_conf,
+ thresh_nms=thresh_nms, in_dim=in_dim, out_path=None,
+ display=display, debug=debug)
+ if debug:
+ max_num_obj = max(max_num_obj, len(obj_list))
for i in range(len(obj_list), max_num_obj):
- cv2.imshow('feature#%d' % i, np.zeros((1, 1), dtype=np.uint8))
- if len(obj_list) > 0:
- cv2.waitKey(0)
+ cv2.imshow('card#%d' % i, np.zeros((1, 1), dtype=np.uint8))
+ if display:
+ cv2.imshow('result', img_result)
+
+ elapsed_ms = (time.time() - start_time) * 1000
+ print('Elapsed time: %.2f ms' % elapsed_ms)
if out_path is not None:
- vid_writer.write(frame.astype(np.uint8))
+ vid_writer.write(img_result.astype(np.uint8))
cv2.waitKey(1)
if out_path is not None:
@@ -253,10 +359,13 @@
def main():
# Specify paths for all necessary files
- test_path = os.path.abspath('../data/test1.jpg')
+ test_path = os.path.abspath('test_file/test4.mp4')
+ #weight_path = 'backup/tiny_yolo_10_39500.weights'
+ #cfg_path = 'cfg/tiny_yolo_10.cfg'
+ #class_path = "data/obj_10.names"
weight_path = 'weights/second_general/tiny_yolo_final.weights'
- cfg_path = 'cfg/tiny_yolo.cfg'
- class_path = "data/obj.names"
+ cfg_path = 'cfg/tiny_yolo_old.cfg'
+ class_path = 'data/obj.names'
out_dir = 'out'
if not os.path.isfile(test_path):
print('The test file %s doesn\'t exist!' % os.path.abspath(test_path))
@@ -271,6 +380,9 @@
print('The class file %s doesn\'t exist!' % os.path.abspath(test_path))
return
+ thresh_conf = 0.01
+ thresh_nms = 0.8
+
# Setup
# Read class names from text file
with open(class_path, 'r') as f:
@@ -290,13 +402,30 @@
if test_ext in ['jpg', 'jpeg', 'bmp', 'png', 'tiff']:
img = cv2.imread(test_path)
- detect_frame(net, classes, img, out_path=out_path)
+ detect_frame(net, classes, img, out_path=out_path, thresh_conf=thresh_conf, thresh_nms=thresh_nms)
else:
- capture = cv2.VideoCapture(test_path)
- detect_video(net, classes, capture, out_path=out_path)
+ capture = cv2.VideoCapture(0)
+ detect_video(net, classes, capture, out_path=out_path, thresh_conf=thresh_conf, thresh_nms=thresh_nms,
+ display=False, debug=False)
capture.release()
pass
if __name__ == '__main__':
+ '''
+ df_list = []
+ for set_name in fetch_data.all_set_list:
+ csv_name = '%s/csv/%s.csv' % (transform_data.data_dir, set_name)
+ df = fetch_data.load_all_cards_text(csv_name)
+ df_list.append(df)
+ #print(df)
+ card_pool = pd.concat(df_list)
+ card_pool.reset_index(drop=True, inplace=True)
+ card_pool.drop('Unnamed: 0', axis=1, inplace=True, errors='ignore')
+ card_pool = calc_image_hashes(card_pool, save_to='card_pool.pck')
+ '''
+ # csv_name = '%s/csv/%s.csv' % (transform_data.data_dir, 'rtr')
+ # card_pool = fetch_data.load_all_cards_text(csv_name)
+ # card_pool = calc_image_hashes(card_pool)
+ card_pool = pd.read_pickle('card_pool.pck')
main()
--
Gitblit v1.10.0