Question in pixel_link.py about the function 'cal_gt_for_single_image()'

[bjxia1]

bbox_masks = []
    pos_mask = mask.copy()
    for bbox_idx, (bbox_xs, bbox_ys) in enumerate(zip(xs, ys)):
        if labels[bbox_idx] == background_label:
            continue
        
        bbox_mask = mask.copy()
        
        bbox_points = zip(bbox_xs, bbox_ys)
        bbox_contours = util.img.points_to_contours(bbox_points)
        util.img.draw_contours(bbox_mask, bbox_contours, idx = -1, 
                               color = 1, border_width = -1)
        
        bbox_masks.append(bbox_mask)
        
        if labels[bbox_idx] == text_label:
            pos_mask += bbox_mask
        
    # treat overlapped in-bbox pixels as negative, 
    # and non-overlapped  ones as positive
    pos_mask = np.asarray(pos_mask == 1, dtype = np.int32)
    num_positive_pixels = np.sum(pos_mask)

Here just draw contours and set contours to 1, but not set inside the contours to 1.
I think here also need use cv2.fillPoly.