import cv2
import face_recognition
import gradio as gr
import datetime as DT
import pytz

ipAddress = None


def __nowInIST():
    return DT.datetime.now(pytz.timezone("Asia/Kolkata"))


def __attachIp(request: gr.Request):
    global ipAddress
    x_forwarded_for = request.headers.get('x-forwarded-for')
    if x_forwarded_for:
        ipAddress = x_forwarded_for


def pprint(log: str):
    now = __nowInIST()
    now = now.strftime("%Y-%m-%d %H:%M:%S")
    print(f"[{now}] [{ipAddress}] {log}")


def __findFaceEncodings(imagePath):
    image = cv2.imread(imagePath)
    faceEncodings = face_recognition.face_encodings(image)
    return faceEncodings[0] if len(faceEncodings) > 0 else None


def predictMatch(firstImage, secondImage):
    image1Encoding = __findFaceEncodings(firstImage)
    image2Encoding = __findFaceEncodings(secondImage)

    pprint("Starting the job...")
    distance = face_recognition.face_distance([image1Encoding], image2Encoding)
    distancePercent = round(distance[0] * 100)
    matchPercent = 100 - distancePercent
    pprint(f"Job finished. Match    : {matchPercent}%")

    isSame = matchPercent > 50

    amplificationFactor = 1.5
    adjustment = (amplificationFactor - 0.5) * 100

    outputTexts = []
    if isSame:
        outputTexts.append("The images are of the same person ✅")
        matchScore = round((matchPercent + adjustment) / (100 + adjustment) * 100)
        outputTexts.append(f"\nMatch Score: {matchScore}%")
    else:
        outputTexts.append("The images are not of the same person ❌")

    outputText = "\n".join(outputTexts)
    pprint(f"{outputText=}")

    return outputText


with gr.Row(elem_classes=["main-container"]):
    with gr.Row(elem_classes=["img-container"]):
        firstImage = gr.Image(type='filepath', height=300, elem_classes=["image"], label="1st Image", container=True)
        secondImage = gr.Image(type='filepath', height=300, elem_classes=["image"], label="2nd Image", container=True)

    with gr.Row(elem_classes=["output-container"]):
        result = gr.Textbox(label="Result", elem_classes=["output"], scale=2)


with gr.Interface(
    fn=predictMatch,
    inputs=[
        firstImage,
        secondImage,
    ],
    outputs=[
        result
    ],
    title="Face Match Detector",
    allow_flagging="never",
) as demo:
    demo.load(__attachIp, None, None)

demo.launch(debug=True)