MISSAOUI's picture
Update app.py
c67fa23 verified
import cv2
import numpy as np
import io
import PIL
from base64 import b64decode, b64encode
from keras.models import load_model
import streamlit as st
from streamlit_webrtc import webrtc_streamer, VideoProcessorBase
# Initialize the Haar Cascade face detection model
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
model = load_model('emotion_model.h5',compile=False)
emotion_dict = {0: "Angry", 1: "Disgust", 2: "Fear", 3: "Happy", 4: "Neutral", 5: "Sad", 6: "Surprised"}
# Define functions to convert between JavaScript image reply and OpenCV image
def js_to_image(js_reply):
image_bytes = b64decode(js_reply.split(',')[1])
jpg_as_np = np.frombuffer(image_bytes, dtype=np.uint8)
img = cv2.imdecode(jpg_as_np, flags=1)
return img
def bbox_to_bytes(bbox_array):
bbox_PIL = PIL.Image.fromarray(bbox_array, 'RGBA')
iobuf = io.BytesIO()
bbox_PIL.save(iobuf, format='png')
bbox_bytes = 'data:image/png;base64,{}'.format((str(b64encode(iobuf.getvalue()), 'utf-8')))
return bbox_bytes
# Define function to process each frame from the video stream
def process_frame(frame):
# Convert frame to grayscale
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# Perform face detection
faces = face_cascade.detectMultiScale(gray)
emotions = []
# Process each detected face
for (x, y, w, h) in faces:
face_region = gray[y:y+h, x:x+w]
face_resized = cv2.resize(face_region, (48, 48))
img = np.expand_dims(face_resized, axis=0)
img = np.expand_dims(img, axis=-1)
predictions = model.predict(img)
emo = model.predict(img)[0]
emotions.append(emo)
predicted_class = np.argmax(predictions)
predicted_emotion = emotion_dict[predicted_class]
accuracy = predictions[0][predicted_class]
# Draw bounding box and emotion label on the frame
cv2.rectangle(frame, (x, y), (x+w, y+h), (255, 0, 0), 2)
cv2.putText(frame, f"{predicted_emotion} ({accuracy:.2f})", (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 0), 2)
return frame, emotions
class VideoProcessor(VideoProcessorBase):
def recv(self, frame):
img = frame.to_ndarray(format="bgr24")
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
for (x, y, w, h) in faces:
face_region = gray[y:y+h, x:x+w]
face_resized = cv2.resize(face_region, (48, 48))
img_array = np.expand_dims(face_resized, axis=0)
img_array = np.expand_dims(img_array, axis=-1)
predictions = model.predict(img_array)
predicted_class = np.argmax(predictions)
predicted_emotion = emotion_dict[predicted_class]
accuracy = predictions[0][predicted_class]
cv2.rectangle(img, (x, y), (x+w, y+h), (255, 0, 0), 2)
cv2.putText(img, f"{predicted_emotion} ({accuracy:.2f})", (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 255, 0), 2)
return frame.from_ndarray(img, format="bgr24")
# Page Title and Description
st.set_page_config(page_title="Facial Emotion Recognition", layout="wide")
st.title("Facial Emotion Recognition")
# Sidebar
st.sidebar.title("Options")
option = st.sidebar.radio("Select Option", ("Drag a File","Process Video"))
# Main Content Area
if option == "Drag a File" :
st.subheader("Photo Processing")
# Process image or captured frame
if option == "Drag a File":
uploaded_file = st.file_uploader("Upload Photo", type=["jpg", "jpeg", "png"])
if uploaded_file is not None:
file_bytes = np.asarray(bytearray(uploaded_file.read()), dtype=np.uint8)
image = cv2.imdecode(file_bytes, 1)
if 'image' in locals():
processed_frame, emotions = process_frame(image)
# Display processed frame and emotions
st.subheader("Processed Frame")
st.image(processed_frame, channels="BGR", use_column_width=False)
if not emotions:
st.warning("No faces detected in the image.")
elif option == "Process Video":
webrtc_streamer(key="camera", video_processor_factory=VideoProcessor)