Danish15/Sign-Language

1import cv22import streamlit as st3import numpy as np4from cvzone.HandTrackingModule import HandDetector5from cvzone.ClassificationModule import Classifier6import math7from PIL import Image8st.set_page_config(page_title="Sign Language Detection", page_icon="🔥",layout="wide")9 10 11# Initialize the camera12cap = cv2.VideoCapture(0)13detector = HandDetector(maxHands=1)14classifier = Classifier("keras_model.h5", "labels.txt")15 16offset = 2017imgSize = 30018 19labels = ["hello", "how are you", "welcome"]20 21# Function to display video stream and detected text22def display_video_with_text():23    st.title("Sign Language Detection")24    st.write("This project employs computer vision and machine learning to recognize and interpret sign language gestures captured via webcam, providing real-time feedback and enabling communication for individuals who use sign language. Through accurate sign language detection, it promotes accessibility and inclusivity in digital interactions and communication platforms.")25    st.write("---")26    effects_images = {27    "hello.jpg": "Hello Sign",28    "how are you.jpg": "How Are You Sign",29    "welcome.jpg": "Welcome Sign"30    }31    st.markdown("### Images:")32    col1, col2, col3 = st.columns(3)33    for i, (image_path, caption) in enumerate(effects_images.items()):34        with col1 if i % 3 == 0 else col2 if i % 3 == 1 else col3:35            st.image(image_path, caption=caption, use_column_width=True, output_format="auto")36    st.write(37        "Place your hand in front of the camera to detect gestures. The detected text will appear on the right side."38    )39    st.write("---")40    41    # Create columns layout42    col1, col2 = st.columns([3, 1])  # Video stream will occupy 3/4 of the width, detected text will occupy 1/443 44    with col1:45        # Video stream placeholder46        st.header("Live Video Streaming:")47        48        video_placeholder = st.empty()49 50    with col2:51        # Detected text placeholder52        st.header("Detected Text:")53        text_placeholder = st.empty()54        55        56 57    while True:58        # Read frame from camera59        ret, img = cap.read()60        if not ret:61            st.error(62                "Failed to retrieve frame from camera. Please check your camera connection."63            )64            break65 66        imgOutput = img.copy()67        hands, img = detector.findHands(img)68        if hands:69            hand = hands[0]70            x, y, w, h = hand["bbox"]71 72            imgWhite = np.ones((imgSize, imgSize, 3), np.uint8) * 25573            imgCrop = img[y - offset : y + h + offset, x - offset : x + w + offset]74 75            imgCropShape = imgCrop.shape76 77            aspectRatio = h / w78 79            # If the cropped image is not empty, resize it80            if imgCropShape[0] != 0 and imgCropShape[1] != 0:81                if aspectRatio > 1:82                    k = imgSize / h83                    wCal = math.ceil(k * w)84                    imgResize = cv2.resize(imgCrop, (wCal, imgSize))85                    imgResizeShape = imgResize.shape86                    wGap = math.ceil((imgSize - wCal) / 2)87                    imgWhite[:, wGap : wCal + wGap] = imgResize88                    prediction, index = classifier.getPrediction(imgWhite)89                else:90                    k = imgSize / w91                    hCal = math.ceil(k * h)92                    imgResize = cv2.resize(imgCrop, (imgSize, hCal))93                    imgResizeShape = imgResize.shape94                    hGap = math.ceil((imgSize - hCal) / 2)95                    imgWhite[hGap : hCal + hGap, :] = imgResize96                    prediction, index = classifier.getPrediction(imgWhite)97            else:98                # If the cropped image is empty, set prediction and index to default values99                prediction, index = "", 0100 101            cv2.putText(102                imgOutput,103                labels[index],104                (x, y - 20),105                cv2.FONT_HERSHEY_COMPLEX,106                2,107                (255, 0, 255),108                2,109            )110 111            # Convert image to RGB for display112            imgOutput = cv2.cvtColor(imgOutput, cv2.COLOR_BGR2RGB)113 114            # Convert image array to PIL Image115            pil_img = Image.fromarray(imgOutput)116 117            # Display video stream and detected text118            video_placeholder.image(pil_img, channels="RGB")119            text_placeholder.write(labels[prediction.index(max(prediction))])120            121 122# Run the Streamlit app123if __name__ == "__main__":124    display_video_with_text()125