Human Detection and Counting
• Introduction
• Brief introduction to the topic.
• Importance of human detection and counting
in various applications like security,
surveillance, crowd management, etc.
• Objectives of the project.
2
• Problem Statement
• Human detection and counting is challenging
due to factors like varied body shapes,
clothing, and movements
• Key difficulties: detecting humans in varied
environments, handling occlusions, lighting
variations.
• Real-world significance : applications in
smart cities, security, retail, healthcare.
3
• Tech Stack
• Programming: Python (OpenCV,
TensorFlow/PyTorch), Flask/Django for
backend.
• Machine Learning Models: YOLO for object
detection, a face recognition model, and an
activity recognition model.
• Hardware: If feasible, you can deploy the
system on an NVIDIA Jetson Nano for
edge processing.
4
• Objective
• Create a surveillance system that can
detect and recognize humans in real time
• Detect and count the number of people in a
specific area using computer vision and
machine learning.
• Identify suspicious activities, and send
alerts.
5
• Features
• Human Detection: Detect people in real-time
using a webcam or video feed.
• Behavior Analysis: Identify specific activities,
like loitering or intrusion.
• Face Recognition: Recognize authorized vs
unauthorized individuals.
• Real-Time Alerts: Send alerts or notifications
when suspicious behavior is detected.
6
Approach and Methodology
•Data Collection & Preprocessing:
• Gather and prepare diverse image/video
data
•Human Detection:
• Use deep learning models to detect humans
•Counting:
• Count people by detecting bounding boxes
around them
7
• Data Collection & Preprocessing
• Gather diverse images/videos from cameras in
different environments, lighting conditions
• Clean the data by removing poor-quality images,
normalize lighting, resize images
• Use techniques like edge detection to enhance
human recognition in complex scenes.
8
• Human Counting
• Counting people based on detected bounding
boxes around each person.
• Using algorithms like SORT to track and count
individuals across video frames.
• Estimating crowd density in crowded areas to count
people accurately.
9
•Results and Evaluation
•Accuracy Metrics: Using Precision, Recall,
and F1-score to
• measure performance.
•Mean Average Precision (map) for object
detection.
•Real-time Performance: Testing for live data
processing
10
Thank you
