Main objective
Development of an automated system for the detection, classification, and counting of people in camera trap images collected within the LIFE LxAquila project “Stewardship Network for the Conservation of Peri-Urban Bonelli’s Eagles”. The main objective is to quantify patterns of human use and assess disturbance levels in areas critical for the conservation of the species, thereby supporting evidence-based management decisions and potential scientific publication.
The work will focus on the design and implementation of a computer vision pipeline based on deep learning models (e.g., object detection and tracking architectures), capable of: (i) automatically detecting people and animals in approximately 1 TB of RGB and infrared images; (ii) estimating the number of individuals per image and per event by aggregating temporal sequences corresponding to the same passage; (iii) classifying the type of use (on foot, vehicle, bicycle, motorcycle, among others); and (iv) exporting structured outputs (CSV format) integrating metadata such as timestamp, event duration, and camera location.
The solution should be robust to challenging conditions (low light, noise, motion blur, variable angles, false positives), scalable to large datasets, and support quantitative performance evaluation, including potential adaptation or fine-tuning of pre-trained models. The final system should be applied to more than 1 TB of images, and results exported in CSV format.
Available data
More than 1 TB of JPEG images collected using 20 motion-triggered camera traps. A substantial subset of images has been manually annotated to enable supervised training of the system.
Proposed tasks
- Development of an automated system for the detection, classification, and counting of people in camera trap images
- Automatic classification and counting of people in 1 TB of JPEG images according to the defined criteria
- Export of the classification and counting results in CSV format
Framework of the topic
- Is the topic part of an ongoing project/study? Yes – funded project
- Is there any financial support? Yes: fieldwork/material
Supervisor/tutor
Rita Ferreira (Terrestrial Conservation Area) | rita.ferreira@spea.pt
