Extract image from video for image labeling/annotating (video-to-image.py) . || Visualize and verify label (visualize-and-verify-label.py)

video-to-image.py

import cv2
import argparse

def Convert_(current, dest_):
    cap = cv2.VideoCapture(current)
    counter=0
    while(True):
        # Capture frame-by-frame
        ret, frame = cap.read()
        if frame is not None:
                if counter%30==0:
                        im = cv2.resize(frame, (640, 360))
                        cv2.imwrite(dest_+str(counter)+'.jpg', frame)
                        print ('saved image no: ', counter)
                counter += 1
        else: break
        
    # When everything done, release the capture
    cap.release()
    cv2.destroyAllWindows()
    
    
if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--src', required=False, dest='src', type=str,
                        default='example.mp4', help='Full path of video')
    parser.add_argument('--dest', required=False, dest='dest', type=str,
                        default='d_34635/', help='Full path of image dir')
    args = parser.parse_args()
    Convert_(args.src, args.dest)
    cv2.destroyAllWindows()

#To run the program
#python vid_to_img.py --src video.mp4 --dest video_folder\

visualize-and-verify-label.py

import os
import cv2
import glob
import numpy as np
import xml.etree.ElementTree as ET


def handle_bad_corners(left, right, top, bottom, im_w, im_h):
    left = np.maximum(5, left)
    top = np.maximum(5, top)
    right = np.minimum(im_w, right)
    bottom = np.minimum(im_h, bottom)    
    return (left, right, top, bottom)


def draw_boxes_and_labels(img, obj_, text_label, box_color=(0, 255, 255)):
    img_h, img_w = img.shape[:2]
    font = cv2.FONT_HERSHEY_SIMPLEX
    font_size = 0.5
    font_color = (0, 0, 0)
    # Draw the object boxes
    left, right, top, bottom = handle_bad_corners(obj_[0], obj_[1], obj_[2], obj_[3], img_w, img_h)
    cv2.rectangle(img, (left, top), (right, bottom), box_color, 4)
    # Draw a filled boxes on top of the bounding box (as the background for the labels)
    left1, top1, right1, _ = handle_bad_corners(left-2, top-20, right+2, bottom, img_w, img_h)
    cv2.rectangle(img, (left1, top1), (right1, top), box_color, -1, 1)
    top2 = 10 if top<25 else top-20
    cv2.putText(img, text_label, (left, top), font, font_size, font_color, 1, cv2.LINE_AA)
    return img


def viz_xml_annotation(path):
    for xml_file in glob.glob(path + '/*.xml'):
        tree = ET.parse(xml_file)
        root = tree.getroot()
        prev_fname = ""
        for member in root.findall('object'):
            fname_ = root.find('filename').text
            if (fname_!=prev_fname):
                prev_fname = fname_
            class_ = member[0].text
           
            xmin, ymin = (int(member[4][0].text), int(member[4][1].text))
            xmax, ymax = (int(member[4][2].text), int(member[4][3].text))
            print (fname_, class_)
            img = cv2.imread(os.path.join(path, fname_))
            h = img.shape[0]
            w = img.shape[1]
            crop = 4.0
            img = cv2.resize(img, (int(w/crop), int(h/crop)))
            cv2.putText(img, fname_, (0+10, int(h/crop)-10), cv2.FONT_HERSHEY_SIMPLEX, 1.0, (255, 255, 255), lineType=cv2.LINE_AA) 
            localized_obj = (int(xmin/crop), int(xmax/crop), int(ymin/crop), int(ymax/crop))
            img = draw_boxes_and_labels (img, localized_obj, class_)
            cv2.imshow("output", img)
            cv2.waitKey(1000)

image_path = "d_003243(37740-38760_35)/"
viz_xml_annotation(image_path)