Python code to extract BBox: OpenCV

Python Code

PyTorch

• Draw bbox while reading image in PIL-Image and display contents in iPython notebook

  • import segments

    from PIL import Image, ImageDraw
    import PIL
    import matplotlib.pyplot as plt
    %matplotlib inline
        
    pil_img = Image.open(act_img_path)
    img_draw = ImageDraw.Draw(pil_img)
    img_draw.rectangle((xmin, ymin, xmax, ymax), outline='red')
    box = (xmin, ymin, xmax, ymax)
    cropped_image = pil_img.crop(box)
    plt.imshow(pil_img)
    # draw only cropped-image
    plt.imshow(cropped_image)
    

  • Ploting image using matplotlib for multiple image in a grid

    # here figsize will be size of the image in a grid; first 2 params are `rows` and `cols` respectively.
    _, axarr = plt.subplots(1,2, figsize=(8, 8))
    image = cv2.imread(act_img_path)
    sh_image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
    test_image = sh_image.copy()
    # this function gets the `face` detected using MTCNN-face detector
    bb_image = get_multiface(test_image)
    
    axarr[0].imshow(sh_image)
    axarr[1].imshow(bb_image)
    

• Draw bbox while reading image in CV2-OpenCV based drawing in iPython notebook

  • import segements

    import cv2
    import matplotlib.pyplot as plt
    %matplotlib inline
    def open_image(filename):
      img = cv2.imread(filename)
      img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
      plt.imshow(img)
    open_image(act_img_path)
    

• Reading Data using Pandas for training

  • For data visualization/analysis

    img_file = os.path.join(txt_file_root_path, 'lifestyleimages.txt')
    pd_data = pd.read_csv(img_file, header=None, names=['origIdPerson','imgPath','age','gender','emotion','recognitionDB',
                                                    'ageAnnot','genderAnnot','emotionAnnot'])
    pd_data.head()
    # Removing rows which have 'other' as gender
    pd_data = pd_data[pd_data.genderAnnot!='OTHER']
    pd_data['is_train'] = np.random.choice(2, size=len(pd_data), p=[0.2, 0.8])
    i = np.random.choice(len(pd_data))
    dfi = pd_data.iloc[i]
    print(f"age:{dfi['ageAnnot']}::genderAnnot:{dfi['genderAnnot']}::emotionAnnot:{dfi['emotionAnnot']}")
    print(f"img name : {dfi['imgPath']}")
    act_img_path = os.path.join(image_root_path, str(dfi["imgPath"]))
    open_image(act_img_path)
    

• Displaying the image with bouding-box read through OpenCV cv2 and drawing the attributes using matplotlib to display it in iPython

  • Consider it that you already have all the bounding-boxes or rather pretend it in a np.array.

    import cv2
    import matplotlib.pyplot as plt
    import matplotlib.patches as patches
    %matplotlib inline
        
    plt.figure()
    fig, ax = plt.subplots(1, figsize=(40,50))
    ax.imshow(orig_image)
        
    for i in range(boxes.size(0)):
        box = boxes[i, :]
        xmin = box[0]
        ymin = box[1]
        xmax = box[2]
        ymax = box[3]
        box_w = xmax-xmin+1
        box_h = ymax-ymin+1
        bbox = patches.Rectangle((xmin, ymin), box_w, box_h, linewidth=2, edgecolor='red', facecolor='none')
        ax.add_patch(bbox)
        #cv2.rectangle(orig_image, (box[0], box[1]), (box[2], box[3]), (255, 255, 0), 4)
        #label = f"""{voc_dataset.class_names[labels[i]]}: {probs[i]:.2f}"""
        label_name = f"{class_names[labels[i]]}: {probs[i]:.2f}"
        # cv2.putText(orig_image, label, (box[0] + 20, box[1] + 40), cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 0, 255),2)
        plt.text(xmin, ymin, s=label_name, color='white', verticalalignment='top',
                bbox={'color': 'brown', 'pad': 0})
        
    plt.axis('off')
    plt.show()