Spaces:
Sleeping
Sleeping
updating app
Browse files- app.py +452 -40
- requirements.txt +2 -1
app.py
CHANGED
|
@@ -1,65 +1,477 @@
|
|
| 1 |
import streamlit as st
|
| 2 |
-
from transformers import pipeline
|
| 3 |
-
from PIL import Image
|
| 4 |
import cv2
|
| 5 |
import numpy as np
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
import torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 7 |
|
| 8 |
# --- App Title and Description ---
|
| 9 |
-
st.title("Real-Time Emotion Detection
|
| 10 |
st.write("""
|
| 11 |
-
This app
|
| 12 |
-
|
| 13 |
""")
|
| 14 |
|
| 15 |
-
# --- Load
|
| 16 |
-
# Cache the model loading so it isn’t reloaded on every app interaction.
|
| 17 |
@st.cache_resource(show_spinner=False)
|
| 18 |
-
def load_emotion_detector():
|
| 19 |
-
|
| 20 |
-
|
|
|
|
| 21 |
return classifier
|
| 22 |
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
-
#
|
| 26 |
-
|
| 27 |
-
|
| 28 |
|
| 29 |
-
# ---
|
| 30 |
-
def
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
if input_method == "Upload an Image":
|
| 40 |
uploaded_file = st.file_uploader("Choose an image file", type=["jpg", "jpeg", "png"])
|
|
|
|
| 41 |
if uploaded_file is not None:
|
| 42 |
-
#
|
| 43 |
image = Image.open(uploaded_file).convert("RGB")
|
| 44 |
-
st.image(image, caption="Uploaded Image", use_column_width=True)
|
| 45 |
-
prediction = predict_emotion(image)
|
| 46 |
-
st.subheader("Prediction:")
|
| 47 |
-
st.write(f"**Emotion:** {prediction['label']}")
|
| 48 |
-
st.write(f"**Confidence:** {prediction['score']:.2f}")
|
| 49 |
|
| 50 |
-
|
| 51 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
picture = st.camera_input("Capture an Image")
|
|
|
|
| 53 |
if picture is not None:
|
| 54 |
-
# Load image from the captured file.
|
| 55 |
image = Image.open(picture).convert("RGB")
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
|
| 62 |
-
|
| 63 |
-
st.
|
| 64 |
-
|
| 65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
|
|
|
|
|
|
| 2 |
import cv2
|
| 3 |
import numpy as np
|
| 4 |
+
import time
|
| 5 |
+
import plotly.graph_objects as go
|
| 6 |
+
from transformers import pipeline
|
| 7 |
+
from PIL import Image
|
| 8 |
import torch
|
| 9 |
+
from collections import deque
|
| 10 |
+
|
| 11 |
+
# Set page config
|
| 12 |
+
st.set_page_config(
|
| 13 |
+
page_title="Real-Time Emotion Detection",
|
| 14 |
+
page_icon="😀",
|
| 15 |
+
layout="wide"
|
| 16 |
+
)
|
| 17 |
|
| 18 |
# --- App Title and Description ---
|
| 19 |
+
st.title("Advanced Real-Time Emotion Detection")
|
| 20 |
st.write("""
|
| 21 |
+
This app detects emotions in real-time using your webcam. It tracks facial expressions continuously
|
| 22 |
+
and provides visual feedback on detected emotions.
|
| 23 |
""")
|
| 24 |
|
| 25 |
+
# --- Load Models ---
|
|
|
|
| 26 |
@st.cache_resource(show_spinner=False)
|
| 27 |
+
def load_emotion_detector(model_name="dima806/facial_emotions_image_detection"):
|
| 28 |
+
"""Load the emotion detection model."""
|
| 29 |
+
with st.spinner(f"Loading emotion detection model ({model_name})..."):
|
| 30 |
+
classifier = pipeline("image-classification", model=model_name)
|
| 31 |
return classifier
|
| 32 |
|
| 33 |
+
@st.cache_resource(show_spinner=False)
|
| 34 |
+
def load_face_detector():
|
| 35 |
+
"""Load the face detector model."""
|
| 36 |
+
with st.spinner("Loading face detection model..."):
|
| 37 |
+
# Load OpenCV's face detector
|
| 38 |
+
face_cascade = cv2.CascadeClassifier(cv2.data.haarcascades + 'haarcascade_frontalface_default.xml')
|
| 39 |
+
return face_cascade
|
| 40 |
+
|
| 41 |
+
# --- Sidebar: Model and Settings ---
|
| 42 |
+
st.sidebar.header("Settings")
|
| 43 |
+
|
| 44 |
+
# Model selection
|
| 45 |
+
model_options = {
|
| 46 |
+
"Facial Emotions (Default)": "dima806/facial_emotions_image_detection",
|
| 47 |
+
"Facial Expressions": "juliensimon/distilbert-emotion"
|
| 48 |
+
}
|
| 49 |
+
selected_model = st.sidebar.selectbox(
|
| 50 |
+
"Choose Emotion Model",
|
| 51 |
+
list(model_options.keys())
|
| 52 |
+
)
|
| 53 |
+
|
| 54 |
+
# Input method selection
|
| 55 |
+
input_method = st.sidebar.radio(
|
| 56 |
+
"Choose Input Method",
|
| 57 |
+
["Real-time Webcam", "Upload an Image", "Capture Image"]
|
| 58 |
+
)
|
| 59 |
+
|
| 60 |
+
# Confidence threshold
|
| 61 |
+
confidence_threshold = st.sidebar.slider(
|
| 62 |
+
"Confidence Threshold",
|
| 63 |
+
min_value=0.0,
|
| 64 |
+
max_value=1.0,
|
| 65 |
+
value=0.5,
|
| 66 |
+
step=0.05
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
# Face detection toggle
|
| 70 |
+
use_face_detection = st.sidebar.checkbox("Enable Face Detection", value=True)
|
| 71 |
+
|
| 72 |
+
# History length for real-time tracking
|
| 73 |
+
if input_method == "Real-time Webcam":
|
| 74 |
+
history_length = st.sidebar.slider(
|
| 75 |
+
"Emotion History Length (seconds)",
|
| 76 |
+
min_value=5,
|
| 77 |
+
max_value=60,
|
| 78 |
+
value=10,
|
| 79 |
+
step=5
|
| 80 |
+
)
|
| 81 |
|
| 82 |
+
# Load the selected model
|
| 83 |
+
classifier = load_emotion_detector(model_options[selected_model])
|
| 84 |
+
face_detector = load_face_detector()
|
| 85 |
|
| 86 |
+
# --- Utility Functions ---
|
| 87 |
+
def detect_faces(image):
|
| 88 |
+
"""Detect faces in an image using OpenCV."""
|
| 89 |
+
# Convert PIL Image to OpenCV format
|
| 90 |
+
if isinstance(image, Image.Image):
|
| 91 |
+
opencv_image = np.array(image)
|
| 92 |
+
opencv_image = opencv_image[:, :, ::-1].copy() # Convert RGB to BGR
|
| 93 |
+
else:
|
| 94 |
+
opencv_image = image
|
| 95 |
+
|
| 96 |
+
# Convert to grayscale for face detection
|
| 97 |
+
gray = cv2.cvtColor(opencv_image, cv2.COLOR_BGR2GRAY)
|
| 98 |
+
|
| 99 |
+
# Detect faces
|
| 100 |
+
faces = face_detector.detectMultiScale(
|
| 101 |
+
gray,
|
| 102 |
+
scaleFactor=1.1,
|
| 103 |
+
minNeighbors=5,
|
| 104 |
+
minSize=(30, 30)
|
| 105 |
+
)
|
| 106 |
+
|
| 107 |
+
return faces, opencv_image
|
| 108 |
|
| 109 |
+
def process_image_for_emotion(image, face=None):
|
| 110 |
+
"""Process image for emotion detection."""
|
| 111 |
+
if isinstance(image, np.ndarray):
|
| 112 |
+
# Convert OpenCV image to PIL
|
| 113 |
+
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
|
| 114 |
+
image = Image.fromarray(image)
|
| 115 |
+
|
| 116 |
+
if face is not None:
|
| 117 |
+
# Crop to face region
|
| 118 |
+
x, y, w, h = face
|
| 119 |
+
image = image.crop((x, y, x+w, y+h))
|
| 120 |
+
|
| 121 |
+
return image
|
| 122 |
|
| 123 |
+
def predict_emotion(image):
|
| 124 |
+
"""Predict emotion from an image."""
|
| 125 |
+
try:
|
| 126 |
+
results = classifier(image)
|
| 127 |
+
return results[0] # Return top prediction
|
| 128 |
+
except Exception as e:
|
| 129 |
+
st.error(f"Error during emotion prediction: {str(e)}")
|
| 130 |
+
return {"label": "Error", "score": 0.0}
|
| 131 |
+
|
| 132 |
+
def draw_faces_with_emotions(image, faces, emotions):
|
| 133 |
+
"""Draw rectangles around faces and label with emotions."""
|
| 134 |
+
img = image.copy()
|
| 135 |
+
|
| 136 |
+
# Define colors for different emotions (BGR format)
|
| 137 |
+
emotion_colors = {
|
| 138 |
+
"happy": (0, 255, 0), # Green
|
| 139 |
+
"sad": (255, 0, 0), # Blue
|
| 140 |
+
"neutral": (255, 255, 0), # Cyan
|
| 141 |
+
"angry": (0, 0, 255), # Red
|
| 142 |
+
"surprise": (255, 165, 0), # Orange
|
| 143 |
+
"fear": (128, 0, 128), # Purple
|
| 144 |
+
"disgust": (0, 128, 128) # Brown
|
| 145 |
+
}
|
| 146 |
+
|
| 147 |
+
# Default color for unknown emotions
|
| 148 |
+
default_color = (255, 255, 255) # White
|
| 149 |
+
|
| 150 |
+
for (x, y, w, h), emotion in zip(faces, emotions):
|
| 151 |
+
# Get color based on emotion (lowercase and remove any prefix)
|
| 152 |
+
emotion_key = emotion["label"].lower().split("_")[-1]
|
| 153 |
+
color = emotion_colors.get(emotion_key, default_color)
|
| 154 |
+
|
| 155 |
+
# Draw rectangle around face
|
| 156 |
+
cv2.rectangle(img, (x, y), (x+w, y+h), color, 2)
|
| 157 |
+
|
| 158 |
+
# Add emotion label and confidence
|
| 159 |
+
label = f"{emotion['label']} ({emotion['score']:.2f})"
|
| 160 |
+
cv2.putText(img, label, (x, y-10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, color, 2)
|
| 161 |
+
|
| 162 |
+
return img
|
| 163 |
+
|
| 164 |
+
# --- Main App Logic ---
|
| 165 |
if input_method == "Upload an Image":
|
| 166 |
uploaded_file = st.file_uploader("Choose an image file", type=["jpg", "jpeg", "png"])
|
| 167 |
+
|
| 168 |
if uploaded_file is not None:
|
| 169 |
+
# Load and display image
|
| 170 |
image = Image.open(uploaded_file).convert("RGB")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 171 |
|
| 172 |
+
col1, col2 = st.columns(2)
|
| 173 |
+
with col1:
|
| 174 |
+
st.image(image, caption="Uploaded Image", use_column_width=True)
|
| 175 |
+
|
| 176 |
+
# Process image
|
| 177 |
+
if use_face_detection:
|
| 178 |
+
faces, opencv_image = detect_faces(image)
|
| 179 |
+
|
| 180 |
+
if len(faces) > 0:
|
| 181 |
+
emotions = []
|
| 182 |
+
for face in faces:
|
| 183 |
+
face_img = process_image_for_emotion(image, face)
|
| 184 |
+
emotions.append(predict_emotion(face_img))
|
| 185 |
+
|
| 186 |
+
# Draw faces with emotions
|
| 187 |
+
result_image = draw_faces_with_emotions(opencv_image, faces, emotions)
|
| 188 |
+
|
| 189 |
+
with col2:
|
| 190 |
+
st.image(result_image, caption="Detected Emotions", channels="BGR", use_column_width=True)
|
| 191 |
+
|
| 192 |
+
# Display predictions
|
| 193 |
+
st.subheader("Detected Emotions:")
|
| 194 |
+
for i, (emotion, face) in enumerate(zip(emotions, faces)):
|
| 195 |
+
if emotion["score"] >= confidence_threshold:
|
| 196 |
+
st.write(f"Face {i+1}: **{emotion['label']}** (Confidence: {emotion['score']:.2f})")
|
| 197 |
+
|
| 198 |
+
# Show confidence bars
|
| 199 |
+
top_emotions = classifier(process_image_for_emotion(image, face))
|
| 200 |
+
labels = [item["label"] for item in top_emotions]
|
| 201 |
+
scores = [item["score"] for item in top_emotions]
|
| 202 |
+
|
| 203 |
+
fig = go.Figure(go.Bar(
|
| 204 |
+
x=scores,
|
| 205 |
+
y=labels,
|
| 206 |
+
orientation='h'
|
| 207 |
+
))
|
| 208 |
+
fig.update_layout(
|
| 209 |
+
title=f"Emotion Confidence - Face {i+1}",
|
| 210 |
+
xaxis_title="Confidence",
|
| 211 |
+
yaxis_title="Emotion",
|
| 212 |
+
height=300
|
| 213 |
+
)
|
| 214 |
+
st.plotly_chart(fig, use_container_width=True)
|
| 215 |
+
else:
|
| 216 |
+
st.warning("No faces detected in the image. Try another image or disable face detection.")
|
| 217 |
+
else:
|
| 218 |
+
# Process the whole image
|
| 219 |
+
prediction = predict_emotion(image)
|
| 220 |
+
st.subheader("Prediction:")
|
| 221 |
+
st.write(f"**Emotion:** {prediction['label']}")
|
| 222 |
+
st.write(f"**Confidence:** {prediction['score']:.2f}")
|
| 223 |
+
|
| 224 |
+
elif input_method == "Capture Image":
|
| 225 |
picture = st.camera_input("Capture an Image")
|
| 226 |
+
|
| 227 |
if picture is not None:
|
|
|
|
| 228 |
image = Image.open(picture).convert("RGB")
|
| 229 |
+
|
| 230 |
+
col1, col2 = st.columns(2)
|
| 231 |
+
with col1:
|
| 232 |
+
st.image(image, caption="Captured Image", use_column_width=True)
|
| 233 |
+
|
| 234 |
+
# Process image
|
| 235 |
+
if use_face_detection:
|
| 236 |
+
faces, opencv_image = detect_faces(image)
|
| 237 |
+
|
| 238 |
+
if len(faces) > 0:
|
| 239 |
+
emotions = []
|
| 240 |
+
for face in faces:
|
| 241 |
+
face_img = process_image_for_emotion(image, face)
|
| 242 |
+
emotions.append(predict_emotion(face_img))
|
| 243 |
+
|
| 244 |
+
# Draw faces with emotions
|
| 245 |
+
result_image = draw_faces_with_emotions(opencv_image, faces, emotions)
|
| 246 |
+
|
| 247 |
+
with col2:
|
| 248 |
+
st.image(result_image, caption="Detected Emotions", channels="BGR", use_column_width=True)
|
| 249 |
+
|
| 250 |
+
# Display predictions
|
| 251 |
+
st.subheader("Detected Emotions:")
|
| 252 |
+
for i, (emotion, face) in enumerate(zip(emotions, faces)):
|
| 253 |
+
if emotion["score"] >= confidence_threshold:
|
| 254 |
+
st.write(f"Face {i+1}: **{emotion['label']}** (Confidence: {emotion['score']:.2f})")
|
| 255 |
+
else:
|
| 256 |
+
st.warning("No faces detected in the image. Try another image or disable face detection.")
|
| 257 |
+
else:
|
| 258 |
+
# Process the whole image
|
| 259 |
+
prediction = predict_emotion(image)
|
| 260 |
+
st.subheader("Prediction:")
|
| 261 |
+
st.write(f"**Emotion:** {prediction['label']}")
|
| 262 |
+
st.write(f"**Confidence:** {prediction['score']:.2f}")
|
| 263 |
|
| 264 |
+
elif input_method == "Real-time Webcam":
|
| 265 |
+
st.subheader("Real-time Emotion Detection")
|
| 266 |
+
st.write("Click 'Start' to begin real-time emotion detection using your webcam.")
|
| 267 |
+
|
| 268 |
+
# Create a placeholder for the webcam feed
|
| 269 |
+
video_placeholder = st.empty()
|
| 270 |
+
|
| 271 |
+
# Create a placeholder for metrics
|
| 272 |
+
metrics_placeholder = st.empty()
|
| 273 |
+
|
| 274 |
+
# Create a placeholder for emotion history chart
|
| 275 |
+
chart_placeholder = st.empty()
|
| 276 |
+
|
| 277 |
+
# Initialize session state for tracking emotions over time
|
| 278 |
+
if 'emotion_history' not in st.session_state:
|
| 279 |
+
st.session_state.emotion_history = {}
|
| 280 |
+
st.session_state.last_update_time = time.time()
|
| 281 |
+
st.session_state.frame_count = 0
|
| 282 |
+
|
| 283 |
+
# Start/Stop button
|
| 284 |
+
start_button = st.button("Start" if 'running' not in st.session_state or not st.session_state.running else "Stop")
|
| 285 |
+
|
| 286 |
+
if start_button:
|
| 287 |
+
st.session_state.running = not st.session_state.get('running', False)
|
| 288 |
+
|
| 289 |
+
# If running, capture and process webcam feed
|
| 290 |
+
if st.session_state.get('running', False):
|
| 291 |
+
try:
|
| 292 |
+
# Open the webcam
|
| 293 |
+
cap = cv2.VideoCapture(0)
|
| 294 |
+
|
| 295 |
+
# Check if webcam opened successfully
|
| 296 |
+
if not cap.isOpened():
|
| 297 |
+
st.error("Could not open webcam. Please check your camera settings.")
|
| 298 |
+
st.session_state.running = False
|
| 299 |
+
else:
|
| 300 |
+
# Create deques for tracking emotions
|
| 301 |
+
emotion_deques = {}
|
| 302 |
+
timestamp_deque = deque(maxlen=30*history_length) # Store timestamps for X seconds at 30fps
|
| 303 |
+
|
| 304 |
+
while st.session_state.get('running', False):
|
| 305 |
+
# Read frame
|
| 306 |
+
ret, frame = cap.read()
|
| 307 |
+
|
| 308 |
+
if not ret:
|
| 309 |
+
st.error("Failed to capture frame from webcam")
|
| 310 |
+
break
|
| 311 |
+
|
| 312 |
+
# Flip the frame horizontally for a more natural view
|
| 313 |
+
frame = cv2.flip(frame, 1)
|
| 314 |
+
|
| 315 |
+
# Increment frame count for FPS calculation
|
| 316 |
+
st.session_state.frame_count += 1
|
| 317 |
+
|
| 318 |
+
# Detect faces
|
| 319 |
+
if use_face_detection:
|
| 320 |
+
faces, _ = detect_faces(frame)
|
| 321 |
+
|
| 322 |
+
if len(faces) > 0:
|
| 323 |
+
# Process each face
|
| 324 |
+
emotions = []
|
| 325 |
+
for face in faces:
|
| 326 |
+
face_img = process_image_for_emotion(frame, face)
|
| 327 |
+
emotions.append(predict_emotion(face_img))
|
| 328 |
+
|
| 329 |
+
# Draw faces with emotions
|
| 330 |
+
frame = draw_faces_with_emotions(frame, faces, emotions)
|
| 331 |
+
|
| 332 |
+
# Update emotion history
|
| 333 |
+
current_time = time.time()
|
| 334 |
+
timestamp_deque.append(current_time)
|
| 335 |
+
|
| 336 |
+
for i, emotion in enumerate(emotions):
|
| 337 |
+
if emotion["score"] >= confidence_threshold:
|
| 338 |
+
face_id = f"Face {i+1}"
|
| 339 |
+
if face_id not in emotion_deques:
|
| 340 |
+
emotion_deques[face_id] = deque(maxlen=30*history_length)
|
| 341 |
+
|
| 342 |
+
emotion_deques[face_id].append({
|
| 343 |
+
"emotion": emotion["label"],
|
| 344 |
+
"confidence": emotion["score"],
|
| 345 |
+
"time": current_time
|
| 346 |
+
})
|
| 347 |
+
else:
|
| 348 |
+
# No faces detected
|
| 349 |
+
pass
|
| 350 |
+
else:
|
| 351 |
+
# Process the whole frame
|
| 352 |
+
pil_image = Image.fromarray(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
|
| 353 |
+
emotion = predict_emotion(pil_image)
|
| 354 |
+
|
| 355 |
+
# Display emotion on frame
|
| 356 |
+
cv2.putText(
|
| 357 |
+
frame,
|
| 358 |
+
f"{emotion['label']} ({emotion['score']:.2f})",
|
| 359 |
+
(10, 30),
|
| 360 |
+
cv2.FONT_HERSHEY_SIMPLEX,
|
| 361 |
+
1,
|
| 362 |
+
(0, 255, 0),
|
| 363 |
+
2
|
| 364 |
+
)
|
| 365 |
+
|
| 366 |
+
# Update emotion history
|
| 367 |
+
current_time = time.time()
|
| 368 |
+
timestamp_deque.append(current_time)
|
| 369 |
+
|
| 370 |
+
if "Frame" not in emotion_deques:
|
| 371 |
+
emotion_deques["Frame"] = deque(maxlen=30*history_length)
|
| 372 |
+
|
| 373 |
+
emotion_deques["Frame"].append({
|
| 374 |
+
"emotion": emotion["label"],
|
| 375 |
+
"confidence": emotion["score"],
|
| 376 |
+
"time": current_time
|
| 377 |
+
})
|
| 378 |
+
|
| 379 |
+
# Calculate FPS
|
| 380 |
+
current_time = time.time()
|
| 381 |
+
time_diff = current_time - st.session_state.last_update_time
|
| 382 |
+
if time_diff >= 1.0: # Update every second
|
| 383 |
+
fps = st.session_state.frame_count / time_diff
|
| 384 |
+
st.session_state.last_update_time = current_time
|
| 385 |
+
st.session_state.frame_count = 0
|
| 386 |
+
|
| 387 |
+
# Update metrics
|
| 388 |
+
with metrics_placeholder.container():
|
| 389 |
+
cols = st.columns(3)
|
| 390 |
+
cols[0].metric("FPS", f"{fps:.1f}")
|
| 391 |
+
cols[1].metric("Faces Detected", len(faces) if use_face_detection else "N/A")
|
| 392 |
+
|
| 393 |
+
# Display the frame
|
| 394 |
+
video_placeholder.image(frame, channels="BGR", use_column_width=True)
|
| 395 |
+
|
| 396 |
+
# Update emotion history chart periodically
|
| 397 |
+
if len(timestamp_deque) > 0 and time_diff >= 0.5: # Update chart every 0.5 seconds
|
| 398 |
+
with chart_placeholder.container():
|
| 399 |
+
# Create tabs for each face
|
| 400 |
+
if len(emotion_deques) > 0:
|
| 401 |
+
tabs = st.tabs(list(emotion_deques.keys()))
|
| 402 |
+
|
| 403 |
+
for i, (face_id, emotion_data) in enumerate(emotion_deques.items()):
|
| 404 |
+
with tabs[i]:
|
| 405 |
+
if len(emotion_data) > 0:
|
| 406 |
+
# Count occurrences of each emotion
|
| 407 |
+
emotion_counts = {}
|
| 408 |
+
for entry in emotion_data:
|
| 409 |
+
emotion = entry["emotion"]
|
| 410 |
+
if emotion not in emotion_counts:
|
| 411 |
+
emotion_counts[emotion] = 0
|
| 412 |
+
emotion_counts[emotion] += 1
|
| 413 |
+
|
| 414 |
+
# Create pie chart for emotion distribution
|
| 415 |
+
fig = go.Figure(data=[go.Pie(
|
| 416 |
+
labels=list(emotion_counts.keys()),
|
| 417 |
+
values=list(emotion_counts.values()),
|
| 418 |
+
hole=.3
|
| 419 |
+
)])
|
| 420 |
+
fig.update_layout(title=f"Emotion Distribution - {face_id}")
|
| 421 |
+
st.plotly_chart(fig, use_container_width=True)
|
| 422 |
+
|
| 423 |
+
# Create line chart for emotion confidence over time
|
| 424 |
+
emotions = list(emotion_data)[-20:] # Get the last 20 entries
|
| 425 |
+
times = [(e["time"] - emotions[0]["time"]) for e in emotions]
|
| 426 |
+
confidences = [e["confidence"] for e in emotions]
|
| 427 |
+
emotion_labels = [e["emotion"] for e in emotions]
|
| 428 |
+
|
| 429 |
+
fig = go.Figure()
|
| 430 |
+
fig.add_trace(go.Scatter(
|
| 431 |
+
x=times,
|
| 432 |
+
y=confidences,
|
| 433 |
+
mode='lines+markers',
|
| 434 |
+
text=emotion_labels,
|
| 435 |
+
hoverinfo='text+y'
|
| 436 |
+
))
|
| 437 |
+
fig.update_layout(
|
| 438 |
+
title=f"Emotion Confidence Over Time - {face_id}",
|
| 439 |
+
xaxis_title="Time (seconds)",
|
| 440 |
+
yaxis_title="Confidence",
|
| 441 |
+
yaxis=dict(range=[0, 1])
|
| 442 |
+
)
|
| 443 |
+
st.plotly_chart(fig, use_container_width=True)
|
| 444 |
+
else:
|
| 445 |
+
st.info(f"No emotion data available for {face_id} yet.")
|
| 446 |
+
else:
|
| 447 |
+
st.info("No emotion data available yet.")
|
| 448 |
+
|
| 449 |
+
# Release the webcam when done
|
| 450 |
+
cap.release()
|
| 451 |
+
|
| 452 |
+
except Exception as e:
|
| 453 |
+
st.error(f"Error during webcam processing: {str(e)}")
|
| 454 |
+
st.session_state.running = False
|
| 455 |
+
else:
|
| 456 |
+
# Display a placeholder image when not running
|
| 457 |
+
placeholder_img = np.zeros((300, 500, 3), dtype=np.uint8)
|
| 458 |
+
cv2.putText(
|
| 459 |
+
placeholder_img,
|
| 460 |
+
"Click 'Start' to begin",
|
| 461 |
+
(80, 150),
|
| 462 |
+
cv2.FONT_HERSHEY_SIMPLEX,
|
| 463 |
+
1,
|
| 464 |
+
(255, 255, 255),
|
| 465 |
+
2
|
| 466 |
+
)
|
| 467 |
+
video_placeholder.image(placeholder_img, channels="BGR", use_column_width=True)
|
| 468 |
+
|
| 469 |
+
# --- Footer ---
|
| 470 |
+
st.markdown("---")
|
| 471 |
+
st.markdown("""
|
| 472 |
+
**Tips for Best Results:**
|
| 473 |
+
- Ensure good lighting for accurate face detection
|
| 474 |
+
- Position your face clearly in the frame
|
| 475 |
+
- Try different emotion models for comparison
|
| 476 |
+
- Adjust the confidence threshold if emotions aren't being detected
|
| 477 |
+
""")
|
requirements.txt
CHANGED
|
@@ -2,4 +2,5 @@ streamlit
|
|
| 2 |
transformers
|
| 3 |
pillow
|
| 4 |
opencv-python
|
| 5 |
-
torch
|
|
|
|
|
|
| 2 |
transformers
|
| 3 |
pillow
|
| 4 |
opencv-python
|
| 5 |
+
torch
|
| 6 |
+
plotly
|