import cv2
import mediapipe as mp
import numpy as np
import torch
import torch.nn as nn
import gradio as gr

# ----------------------------
# Labels
# ----------------------------
GESTURE_LABELS = {
    0: "A", 1: "B", 2: "L", 3: "U", 4: "V", 5: "W",
    6: "Z", 7: "F", 8: "five", 9: "one", 10: "three",
    11: "two", 12: "six", 13: "seven", 14: "eight",
    15: "nine", 16: "ten", 17: "E", 18: "four",
    19: "i", 20: "k", 21: "r", 22: "zero",
    23: "m", 24: "s"
}
CONF_THRESHOLD = 0.6

# ----------------------------
# Model
# ----------------------------
class GestureNet(nn.Module):
    def __init__(self, input_size=126, num_classes=len(GESTURE_LABELS)):
        super().__init__()
        self.fc1 = nn.Linear(input_size, 256)
        self.fc2 = nn.Linear(256, 128)
        self.fc3 = nn.Linear(128, num_classes)
        self.relu = nn.ReLU()
        self.dropout = nn.Dropout(0.3)

    def forward(self, x):
        x = self.relu(self.fc1(x))
        x = self.dropout(x)
        x = self.relu(self.fc2(x))
        x = self.dropout(x)
        return self.fc3(x)

model = GestureNet()
model.load_state_dict(torch.load("gesture_model1.pth", map_location="cpu"))
model.eval()

# ----------------------------
# MediaPipe
# ----------------------------
mp_hands = mp.solutions.hands

# ----------------------------
# Predict
# ----------------------------
def predict(image):
    # Gradio sends RGB, convert to BGR for OpenCV then back to RGB for MediaPipe
    image_bgr = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
    image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)

    with mp_hands.Hands(static_image_mode=True, max_num_hands=2) as hands:
        results = hands.process(image_rgb)

    coords = []
    if results.multi_hand_landmarks:
        for hand_landmarks in results.multi_hand_landmarks:
            hand_coords = np.array([[lm.x, lm.y, lm.z] for lm in hand_landmarks.landmark])
            hand_coords -= hand_coords[0]
            max_val = np.max(np.linalg.norm(hand_coords, axis=1))
            if max_val > 0:
                hand_coords /= max_val
            coords.extend(hand_coords.flatten())

    if len(coords) < 126:
        coords.extend([0.0] * (126 - len(coords)))
    elif len(coords) > 126:
        coords = coords[:126]

    if len(coords) == 126:
        input_tensor = torch.tensor(coords, dtype=torch.float32).unsqueeze(0)
        with torch.no_grad():
            outputs = model(input_tensor)
            probs = torch.softmax(outputs, dim=1)
            pred_class = torch.argmax(probs, dim=1).item()
            confidence = probs[0][pred_class].item()
        if confidence >= CONF_THRESHOLD:
            return f"{GESTURE_LABELS[pred_class]} ({confidence*100:.2f}%)"

    return "Unknown"

# ----------------------------
# Gradio UI
# ----------------------------
app = gr.Interface(
    fn=predict,
    inputs=gr.Image(type="numpy"),
    outputs="text",
    title="Hand Gesture Recognition",
    description="Upload an image of a hand gesture to recognize it."
)
app.launch()