app.py

import numpy as np
from PIL import Image
import gradio as gr
from ultralytics import YOLO
import cv2
import os

# Load the YOLO model
m_raw_model = YOLO("M-Raw.pt")
n_raw_model = YOLO("N-Raw.pt")
s_raw_model = YOLO("S-Raw.pt")
m_pre_model = YOLO("M-Pre.pt")
n_pre_model = YOLO("N-Pre.pt")
s_pre_model = YOLO("S-Pre.pt")

# Class to syllable map
class_mapping = {0: 'Baybayin Character', 1: 'a', 2: 'b', 3: 'ba', 4: 'be', 5: 'bi', 6: 'bo', 7: 'bu', 8: 'd', 9: 'da', 10: 'di', 11: 'do', 12: 'du', 13: 'e', 14: 'g', 15: 'ga', 16: 'gi', 17: 'go', 18: 'gu', 19: 'ha', 20: 'he', 21: 'hi', 22: 'ho', 23: 'hu', 24: 'i', 25: 'k', 26: 'ka', 27: 'ki', 28: 'ko', 29: 'ku', 30: 'l', 31: 'la', 32: 'le', 33: 'li', 34: 'lo', 35: 'lu', 36: 'm', 37: 'ma', 38: 'me', 39: 'mi', 40: 'mo', 41: 'mu', 42: 'n', 43: 'na', 44: 'ng', 45: 'nga', 46: 'ngi', 47: 'ngo', 48: 'ngu', 49: 'ni', 50: 'no', 51: 'nu', 52: 'o', 53: 'p', 54: 'pa', 55: 'pe', 56: 'pi', 57: 'po', 58: 'pu', 59: 'r', 60: 'ra', 61: 're', 62: 'ri', 63: 'ro', 64: 'ru', 65: 's', 66: 'sa', 67: 'se', 68: 'si', 69: 'so', 70: 'su', 71: 't', 72: 'ta', 73: 'te', 74: 'ti', 75: 'to', 76: 'tu', 77: 'u', 78: 'w', 79: 'wa', 80: 'we', 81: 'wi', 82: 'y', 83: 'ya', 84: 'yi', 85: 'yo', 86: 'yu'}

#-------- 
# Some text
#--------

pageTitle = "Baybayin Instance Detection"
msgWarning = "This demo was created by Adriel Amoguis and Miguel Flores for PCSC 2023"
desc = """
<center>
Take a picture of some Baybayin text and upload it in the Image Box on the left. You can choose the confidence threshold and the IoU threshold using the sliders. <br />
<img src="https://qph.cf2.quoracdn.net/main-qimg-a5ae0e74f5d0ef9980e7b1c59fdf2014-lq" width="60%">
</center>
"""


def snap(upload, conf, iou, show_conf=False):

    # if webcam is not None:
    #     image = webcam
    # elif upload is not None:
    #     image = upload
    # else:
    #     image = webcam
    image = upload

    # If no model selected, use M-Raw
    # if model == None:
    model = "M-Raw"
    
    # Run the selected model
    results = None
    if model == "M-Raw":
        results = m_raw_model(image, conf=conf, iou=iou)
    elif model == "N-Raw":
        results = n_raw_model(image, conf=conf, iou=iou)
    elif model == "S-Raw":
        results = s_raw_model(image, conf=conf, iou=iou)
    elif model == 'M-Pre':
        results = m_pre_model(image, conf=conf, iou=iou)
    elif model == 'N-Pre':
        results = n_pre_model(image, conf=conf, iou=iou)
    elif model == 'S-Pre':
        results = s_pre_model(image, conf=conf, iou=iou)

    # Convert the results list into an output image
    result = results[0]

    if result.boxes.cls.cpu().numpy().size == 0:
        print("No detections.")
        return image
    
    
    classes = result.boxes.cls.cpu().numpy()
    probs = result.boxes.conf.cpu().numpy()
    boxes = result.boxes.xyxy.cpu().numpy()

    # print(classes)
    # print(probs)
    # print(boxes)

    # print(f"Detected {classes} with {probs:.2f} confidence.")
    words = []
    for i in range(len(boxes)):
        x1, y1, x2, y2 = boxes[i]
        x1, y1, x2, y2 = int(x1), int(y1), int(x2), int(y2)
        cv2.rectangle(image, (x1, y1), (x2, y2), (255, 0, 0), int(image.shape[1] * 0.001))
        if show_conf: cv2.putText(image, f"{class_mapping[int(classes[i])]} {probs[i]:.2f}", (x1, y1), cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 0, 0), int(image.shape[1] * 0.001))
        else: cv2.putText(image, f"{class_mapping[int(classes[i])]}", (x1, y1), cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 0, 0), int(image.shape[1] * 0.00075))
        # words.append([(x1,y1), (x2,y2), class_mapping[int(classes[i])]])

    # Sort the words by top to bottom, left to right
    # words = sorted(words, key=lambda x: (x[0][1], x[0][0]))
    # words = [word[2] for word in words]

    return image
            
# Get the samples array
# samples_dir = "sample_images"
# samples = os.listdir(samples_dir)
# samples = [os.path.join(samples_dir, sample) for sample in samples]

demo = gr.Interface(
    snap,
    [
    # gr.Webcam(type="numpy", label="Webcam"), 
     gr.Image(source="upload", type="numpy", label="Baybayin Image"),
    #  gr.Radio(["M-Raw", "S-Raw", "N-Raw", "M-Pre", "S-Pre", "N-Pre"], label="Model", value="M-Raw"), 
     gr.Slider(0, 1, value=0.5, label="Classifier Confidence Threshold"), 
     gr.Slider(0, 1, value=0.5, label="IoU Threshold")],
    #  gr.Checkbox(label="Show Confidence Scores", value=False),
    [gr.Image(type="numpy", label="Detected Baybayin")], #gr.Textbox(lines=5, label="Detected Latin Syllabes")],
    title="Baybayin Instance Detection",
    article=msgWarning,
    description=desc
).queue()

if __name__ == "__main__":
    demo.launch()