app.py

import gradio as gr
import cv2
import numpy as np
from PIL import Image
import pickle
import tensorflow as tf
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing.image import img_to_array
import easyocr
import torch

# ========== GPU Checks ==========
print("Torch GPU Available:", torch.cuda.is_available())
print("TensorFlow GPU Devices:", tf.config.list_physical_devices('GPU'))

# ========== Load Model and Label Encoder ==========
model_path = "MobileNetBest_Model.h5"
label_path = "MobileNet_Label_Encoder.pkl"

model = load_model(model_path)
print("✅ MobileNet model loaded.")

# Label encoder
try:
    with open(label_path, 'rb') as f:
        label_map = pickle.load(f)
        index_to_label = {v: k for k, v in label_map.items()}
    print("✅ Label encoder loaded:", index_to_label)
except:
    index_to_label = {0: "Handwritten", 1: "Computerized"}
    print("⚠️ Default labels used:", index_to_label)

# ========== Initialize EasyOCR (Force GPU) ==========
reader = easyocr.Reader(['en'], gpu=torch.cuda.is_available())
print("✅ EasyOCR initialized with GPU:", torch.cuda.is_available())

# ========== Classify One Region ==========
def classify_text_region(region_img):
    try:
        region_img = cv2.resize(region_img, (224, 224))
        region_img = region_img.astype("float32") / 255.0
        region_img = img_to_array(region_img)
        region_img = np.expand_dims(region_img, axis=0)

        preds = model.predict(region_img)

        if preds.shape[-1] == 1:
            return "Computerized" if preds[0][0] > 0.5 else "Handwritten"
        else:
            class_idx = np.argmax(preds[0])
            return index_to_label.get(class_idx, "Unknown")
    except Exception as e:
        print("❌ Classification error:", e)
        return "Unknown"

# ========== OCR & Annotate ==========
def AnnotatedTextDetection_EasyOCR_from_array(img):
    results = reader.readtext(img)
    annotated_results = []

    for (bbox, text, conf) in results:
        if conf < 0.3 or text.strip() == "":
            continue

        x1, y1 = map(int, bbox[0])
        x2, y2 = map(int, bbox[2])
        crop = img[y1:y2, x1:x2]

        if crop.size == 0:
            continue

        label = classify_text_region(crop)
        annotated_results.append(f"{text.strip()} → {label}")

        color = (0, 255, 0) if label == "Computerized" else (255, 0, 0)
        cv2.rectangle(img, (x1, y1), (x2, y2), color, 2)
        cv2.putText(img, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 0.6, color, 1)

    return cv2.cvtColor(img, cv2.COLOR_BGR2RGB), "\n".join(annotated_results)

# ========== Inference Function ==========
def infer(image):
    img = np.array(image)

    max_dim = 1000
    if img.shape[0] > max_dim or img.shape[1] > max_dim:
        scale = max_dim / max(img.shape[0], img.shape[1])
        img = cv2.resize(img, (int(img.shape[1]*scale), int(img.shape[0]*scale)))

    annotated_img, result_text = AnnotatedTextDetection_EasyOCR_from_array(img)
    return Image.fromarray(annotated_img), result_text

# ========== Gradio UI ==========
with gr.Blocks(
    title="Text Type Classifier",
    css="""
        body {
            background-color: #C0C0C0 !important;
        }
        .outer-box {
            border: 8px solid black;
            border-radius: 16px;
            padding: 24px;
            background-color: white;
        }
        .gr-box {
            border: 6px solid #0288d1 !important;
            border-radius: 12px;
            padding: 16px;
            background-color: white;
            box-shadow: 0px 2px 10px rgba(0,0,0,0.1);
        }
        .gr-button {
            background-color: #0288d1 !important;
            color: white !important;
            font-weight: bold;
            border-radius: 8px;
            margin-top: 10px;
        }
        .gr-button:hover {
            background-color: #01579b !important;
        }
    """
) as demo:

    with gr.Column(elem_classes=["outer-box"]):

        gr.Markdown(
            """
            <div style="text-align: center;">
                <h1><strong>Handwritten vs Computerized Text Classifier</strong></h1>
            </div>
            """,
            elem_classes=["gr-box"]
        )

        with gr.Row():
            with gr.Column():
                image_input = gr.Image(label="Upload Image", type="numpy", elem_classes=["gr-box"])
                submit_btn = gr.Button("Process Image", elem_classes=["gr-box", "gr-button"])
                clear_btn = gr.Button("Clear", elem_classes=["gr-box", "gr-button"])

            with gr.Column():
                image_output = gr.Image(label="Annotated Output", type="numpy", elem_classes=["gr-box"])
                text_output = gr.Textbox(label="Detected Results", lines=10, elem_classes=["gr-box"])

        submit_btn.click(
            fn=infer,
            inputs=image_input,
            outputs=[image_output, text_output]
        )

        clear_btn.click(
            fn=lambda: (None, None, ""),
            inputs=[],
            outputs=[image_input, image_output, text_output]
        )

demo.launch()