This tool combines our own and well-known camera trap image classification pre-trained models under one roof. It integrates them via an ensemble model to increase the trustworthiness of the classification decision. It aims to put conservation biologists at ease as multiple models process each image to detect animals, the number of animals, types and the activities they perform.

This project examines the optimization of deep learning models for edge devices, focusing on Convolutional Neural Networks (CNNs) using acoustic data. It compares seven CNNs for data augmentation, feature extraction, and model compression. Key findings include achieving optimal balance between accuracy and size, especially notable in MobileNet-v3-small and ACDNet, with accuracies of 87.95% and 85.64%, and sizes of 243 KB and 484 KB, respectively. The study proves CNNs' effectiveness in compact, resource-limited environments like forest sound classification.

The study of environmental sound classification (ESC) has become popular over theyears due to the intricate nature of environmental sounds and the evolution of deep learning (DL) techniques. Forest ESC is one use case of ESC, which has been widely experimented with recently to identify illegal activities inside a forest. However, at present, there is a limitation of public datasets specific to all the possible sounds in a forest environment. FSC22 fills the gap of a benchmark dataset for forest environmental sound classification.

ForestQB is a SPARQL query builder to assist Bioscience and Wildlife Researchers in accessing Linked Data. As they are unfamiliar with the Semantic Web and the data ontologies, ForestQB aims to empower them to benefit from using Linked-Data to extract valuable information without grasping the data's nature and underlying technologies. ForestQB is integrating Form-Based Query builders with Natural Language to simplify query construction and match user requirements.

Observing forest environments gives researchers vital information for effective planning and appropriate responses to these events. It can also help spotlight trends that otherwise may not have been noticed. This project explores various data science techniques to tell a story of how the Kinabatangan Forest and the surrounding area of Sabah have changed over time. 

The project explores the dataset with the intention of helping patrollers better understand elephant movement and thus locate them easily. With the prevalence of consistent elephant movement trends in the usage of natural forest corridors, this project aims to develop and train a machine learning regression framework to predict the future GPS locations and movement trends of a Bornean pygmy elephant located in Sabah, Malaysia, in the hopes that it can be used as a secondary tool to base patrol routes around. 

A linked data knowledge graph powers this platform. It integrates heterogeneous data types such as animal collar data, camera traps, satellite imagery, LiDAR and environmental data. Each data set is generated using different time frames, durations, geographic areas. It facilitates non-technical users to discover data and insights using web and conversational AI user interfaces. It also comprises data visualisations capabilities.

This is a toolkit comprised of both hardware and software components. It allows creating low power long-range network using Near Vertical Incidence skywave (NVIS) to facilitate reliable communication in remote jungle environments.

This is a tool built by extending Traccar and open source software  GPS tracking server. We extended its capability by developing and integrating an SMS server that enables us to track using RPS nodes that can send SMS messages.

This is a tool specifically implemented for low resource edge devices and performed well for extreme illumination changes such as day and nighttime. The accuracy, energy efficiency, communication and computational latency of the proposed models are validated using different license plate datasets in the daytime and nighttime and in real-time.