face_recognition/helper.py

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

def show_pred_image(tree,img):
    root=tree.getroot()

    size=root.find('size')
    w,h=int(size.find("width").text),int(size.find("height").text)
    default_color=[220,255,0]
    random_color=np.random.randn(3)*255

    for obj in root.findall("object"):
        bndbox=obj.find("bndbox")
        classname=obj.find('name').text

        xmin,ymin , xmax,ymax=int(bndbox.find('xmin').text),int(bndbox.find('ymin').text),int(bndbox.find('xmax').text),int(bndbox.find('ymax').text)

        if obj.find("distance") is None:
            color=default_color  
        else:
            color=random_color
            distance=obj.find("distance").text
            classname+=f"({distance})"

        img=cv2.rectangle(img,(xmin,ymin),(xmax,ymax),color,7)
        img=cv2.putText(img,classname,(xmin,ymin),cv2.FONT_HERSHEY_SIMPLEX,2,color,thickness=5)
    return img


def xml_to_objs_found(tree):
    root=tree.getroot()

    size=root.find('size')
    image_w,image_h=int(size.find("width").text),int(size.find("height").text)
    
    objs_found=[]
    for obj in root.findall("object"):
        obj_found=dict()
        bndbox=obj.find("bndbox")
        classname=obj.find('name').text
        distance=float(obj.find('distance').text) if classname!="face" else None

        xmin,ymin , xmax,ymax=int(bndbox.find('xmin').text),int(bndbox.find('ymin').text),int(bndbox.find('xmax').text),int(bndbox.find('ymax').text)
        
        x,y = xmin/image_w , ymin/image_h
        w,h = (xmax-xmin)/image_w , (ymax-ymin)/image_h

        obj_details={'xywh':[x,y,w,h],'class':classname,'distance':distance}
        objs_found.append(obj_details)

    return objs_found


def objs_found_to_xml(test_img,w,h,objs_found):

    def rescale(obj_found,w,h):
        # xywh
        obj_found[0]*=w
        obj_found[1]*=h
        obj_found[2]*=w
        obj_found[3]*=h
        return obj_found

    root=ET.Element("annotation")

    filename_tag=ET.Element("filename")
    filename_tag.text=test_img
    root.append(filename_tag)
    path_tag=ET.Element("path")
    path_tag.text="./"+test_img
    root.append(path_tag)


    size_tag=ET.Element("size")
    # w,h defined above
    # print(w,h)
    width_tag=ET.Element("width")
    width_tag.text=str(w)
    height_tag=ET.Element("height")
    height_tag.text=str(h)
    depth_tag=ET.Element("depth")
    depth_tag.text="3"

    size_tag.append(width_tag)
    size_tag.append(height_tag)
    size_tag.append(depth_tag)
    root.append(size_tag)

    # add all objects
    for obj_found in objs_found:
        obj_found[2:]=rescale(obj_found[2:],w,h)

        obj_tag=ET.Element("object")
        name_tag=ET.Element("name")
        name_tag.text=obj_found[1]
        obj_tag.append(name_tag)

        bndbox_tag=ET.Element("bndbox")
        xmin_tag=ET.Element("xmin")
        xmin_tag.text=str(int(obj_found[2]))
        ymin_tag=ET.Element("ymin")
        ymin_tag.text=str(int(obj_found[3]))
        xmax_tag=ET.Element("xmax")
        xmax_tag.text=str(int(obj_found[2]+obj_found[4]))
        ymax_tag=ET.Element("ymax")
        ymax_tag.text=str(int(obj_found[3]+obj_found[5]))

        bndbox_tag.append(xmin_tag)
        bndbox_tag.append(ymin_tag)
        bndbox_tag.append(xmax_tag)
        bndbox_tag.append(ymax_tag)

        obj_tag.append(bndbox_tag)
        root.append(obj_tag)
        
    xml=ET.ElementTree(root)

    # with open(xml_file_path,"wb") as f:
    #     xml.write(f)

    return xml