Computer Vision · Data Visualisation · Team Project
Endangered Animal Recognition
A camera-based species classification application built with Python, OpenCV, Keras and Streamlit, combining confidence-based filtering with timestamped prediction logs and an analytics dashboard.
Technology Stack
Overview
This team project connects camera input to an image classifier and presents predictions through a Streamlit interface. It records timestamped species predictions and confidence values for later analysis.
How it works
Frames are resized to 224 × 224 pixels and normalised before Keras inference. Confidence filtering determines which predictions are logged, while a cooldown limits repeated entries for the same species. A dashboard summarises recorded observations using Pandas and Plotly.
- Camera
- Preprocessing
- Keras classifier
- Confidence filter
- Prediction log
- Dashboard
Scope
The application classifies the image as a whole. It does not provide bounding-box object detection or verified multi-animal tracking. Model confidence controls logging; it is not a measured test-accuracy score.
Role & Scope
A team project in which I contributed to the team’s camera-based classification application. The classifier is a pre-trained Keras model that the app loads.