This website uses strictly required cookies for authentication and secure processing of user data. It does not use trackers or collect any information beyond what users explicitly choose to store. An exception to this rule are the cookies of streaming services (e.g. YouTube) which are loaded on pages with embedded videos. By continuing to use the site, you agree to this cookie policy.
Home
MegaDetector Video Counting
×
MegaDetector example
MegaDetector example
Image source: Sadoun © 2025 All rights reserved

Adapted MegaDetector algorithm (AI model that identifies animals, people and vehicles in camara trap images) to count animals, people and vehicles in short videos.

Key characteristics

Work step
Data analysis
Tool purpose
Photo/Video Data
Properties
Experimental
Keywords
HUMANITA
Trail
Visitor monitoring
Counting people
Camera

Tool description

MegaDetector is an open-source, AI-based object detection system hosted on GitHub and developed by Microsoft. It is designed to automatically identify animals, people, and vehicles in images. It does not identify vehicle classes, but instead flags images for further review or downstream analysis. A further development of MegaDetector allows to count animals, people and vehicles in short video clips. It doesn't track over time but records the peak number of animals, people and vehicles per video. Additionally, it pixelizes people to encounter data privacy regulations. Enabling automatic counting from motion-triggered video clips reduces manual (human) effort, thus reducing operation cost.

Constraints

  • It does not identify species or vehicle types, but instead flags images/videos for further review or downstream analysis.
  • Not the total number of passing objects but the peak number of classes per video is counted which may lead to undercounting.
  • The same object might appear on the 2 subsequent videos which may lead to a double counting.
  • Privacy and data protection constraints may limit deployment and data storage

Requirements

  • Python
  • Trail camera with continuous power supply
  • Computing infrastructure for data processing (local or cloud-based)

Tool Impact

AI-based photo and video analysis may raise privacy concerns related to the recording of visitors, even when data are anonymized or processed automatically. The installation and maintenance of cameras can cause localized disturbance to vegetation and fauna, particularly if repeated site visits are required.

Best Practices

  •  Within Interreg CE HUMANITA and Interreg It-At INDIALPS project the Megadetector algorithm was tested and further developed to enable animal, people and vehicle counting in short video clips.

Helpful hints to use the tool proficiently

  • Make sure the camera view covers a sufficient street length preventing the vehicles from appearing only at frame ends.
  • Place the camera on a curve where vehicles slow down, giving the motion triggered camera enough time to switch on and off.
  • Make sure the power supply to the camera is continuous preventing the camera from going off at peak traffic times.

Specification

Category Software
Platform
Desktop
Operating system
Windows
ModeOnly offline

Linked tools

Category Tool title and description
Study object
Study focus
Work step
Tool purpose
 Classic Professional Free to use Experimental
AI-based visitor counting from photo and video data

Use of AI-based image and video analysis software to automatically detect and count visitors from camera data, supporting continuous and non-intrusive monitoring of visitor numbers and flows in protected areas.
Automatic visitor counters

Automated sensor-based systems for quantifying visitor numbers and analyzing temporal and spatial visitation patterns in natural areas.
Camera traps for visitor monitoring

Passive recording of human presence along trails and access points using motion-triggered camera traps.
Camera traps for wildlife monitoring

Motion-activated camera trap for remote, non-invasive wildlife and visitor monitoring providing valuable data on species presence, abundance, activity patterns, and behaviour over time.

Gallery

The gallery is empty.

Legend

Study objects

Microbiota
Humans
Animals
Vegetation, Funghi & Lichen
Habitats

Study focus

Anthropogenic factors and impacts
Biodiversity (Species populations and Community Composition)
Ecosystem State, Structure and Functioning

Work steps

Presentation
Data collection
Plot establishment
Data analysis
Data archiving

Tool categories

TBD
Device
Software
Manual / Guideline / Framework

Tool purposes

Spatial Data
Numeric and Alphanumeric Data
Audio Data
Genetic Data
Photo/Video Data
Non Data generative
Chemical Compound Data