Initial commit for Face Counter & Density Estimator app

README.md

@@ -0,0 +1,63 @@
+🎯 Project Title: Face Counting and Crowd Density Estimation using MTCNN
+🔍 Overview
+This project addresses the problem of estimating the number of faces in an image, with the added goal of classifying crowd density as Sparse, Medium, or Dense. Initially approached as a regression task, we shifted strategies after empirical evaluation showed that using ground-truth bounding box annotations from the WIDER FACE dataset provided more reliable face counts than a trained regressor. We ultimately adopted the MTCNN (Multi-task Cascaded Convolutional Networks) approach for real-time face detection due to its robustness, accuracy, and speed in varying crowd scenarios.
+
+🧠 Why MTCNN?
+MTCNN is a popular face detection framework because:
+
+It combines face detection and facial landmark localization, making it suitable for fine-grained analysis.
+
+It performs well across scales, poses, and lighting conditions.
+
+It runs in real-time, enabling its use in live applications like webcam feeds.
+
+It's pretrained and optimized, saving time and training resources.
+
+💡 How It Works
+A user uploads an image or uses their webcam to capture one.
+
+The app runs MTCNN to detect faces and count bounding boxes.
+
+Based on the count:
+
+Sparse: 1–10 faces
+
+Medium: 11–50 faces
+
+Dense: 51+ faces
+
+Results are overlaid directly on the image, offering a visual interpretation of the density.
+
+🛠️ Tech Stack
+Model: facenet-pytorch MTCNN
+
+Dataset: WIDER FACE (for validation and benchmarking)
+
+Framework: Python, Streamlit
+
+Libraries: OpenCV, PIL, Torch, Matplotlib
+
+Deployment: Streamlit (local or cloud-hosted)
+
+🌍 Real-Life Applications
+Surveillance & Public Safety
+Detect unusually dense crowds in public areas to trigger alerts.
+
+Event Management
+Monitor real-time foot traffic and optimize crowd control in concerts, rallies, etc.
+
+Retail Analytics
+Gauge customer distribution across zones in malls or stores.
+
+Transportation Hubs
+Analyze crowd density in airports or stations to deploy personnel dynamically.
+
+Smart Cities
+Integrated with CCTV systems, the app can be part of intelligent urban monitoring.
+
+Pandemic Safety Enforcement
+Identify when crowd limits are exceeded to enforce health protocols.
+
+✅ Impact
+This project demonstrates how a lightweight, pretrained architecture like MTCNN can replace heavyweight regression-based models when data annotation is reliable. It bridges computer vision with social and safety applications, showing how academic tools can translate into real-world solutions.
+

app.py

@@ -0,0 +1,57 @@
+import gradio as gr
+from PIL import Image, ImageDraw
+from facenet_pytorch import MTCNN
+import torch
+
+# Load MTCNN model
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+mtcnn = MTCNN(keep_all=True, device=device)
+
+# Face detection and density classification function
+def detect_faces(image):
+    if image is None:
+        return None, "No image provided.", "No density calculated."
+    
+    # Detect faces
+    boxes, _ = mtcnn.detect(image)
+    face_count = 0 if boxes is None else len(boxes)
+
+    # Classify density
+    if face_count <= 10:
+        density = "🟢 Sparse"
+    elif face_count <= 50:
+        density = "🟡 Medium"
+    else:
+        density = "🔴 Dense"
+
+    # Annotate image
+    annotated = image.copy()
+    draw = ImageDraw.Draw(annotated)
+    if boxes is not None:
+        for box in boxes:
+            draw.rectangle(box.tolist(), outline="red", width=3)
+
+    # Return annotated image and stats
+    return annotated, f"🧮 Face Count: {face_count}", f"📊 Crowd Density: {density}"
+
+# Gradio UI
+title = "🎯 Face Counter & Density Estimator"
+description = """
+Upload an image or use your webcam to detect faces and estimate crowd density.
+"""
+
+iface = gr.Interface(
+    fn=detect_faces,
+    inputs=gr.Image(sources=["upload", "webcam"], type="pil", label="Upload Image or use Webcam"),
+    outputs=[
+        gr.Image(type="pil", label="Detected Faces"),
+        gr.Textbox(label="Face Count"),
+        gr.Textbox(label="Crowd Density"),
+    ],
+    title=title,
+    description=description,
+    allow_flagging="never",
+    theme="soft"
+)
+
+iface.launch()

requirement.txt

@@ -0,0 +1,4 @@
+gradio
+torch
+facenet-pytorch
+Pillow