Paul Fergus, a professor of machine learning, and Carl Chambers, a reader on the course, explain what you can expect from the Masters, which utilises their Conservation AI platform.
What is Conservation AI?
Conservation AI is a platform we've created that allows us to use AI to analyse the vast quantities of data being generated from conservation practices. We focus on detecting and classifying animals, humans, and man-made objects indicative of poaching. To do this, we use acoustic sensors that are in the wild and thermal infrared cameras that are used on drones or camera traps. We've got ongoing projects in North America and with The Snow Leopard Trust in Asia, we're actually the first organisation to detect a pangolin in real time, using AI in Uganda.
Can you tell us about the real-life projects students can get involved with?
Many of the technologies used in Conservation AI, such as computer vision and large language models (LLMs), that detect different animal species and generate biodiversity insight reports, are taught in the degree.
In terms of coursework, the students have to train their own models to detect different animals species. For the reporting side, they use state of the art LLMs (such as Llama 2 7b) and retrieval augmented generation (RAG) to process data and generate reports in natural language.
Students have also developed audio machine learning algorithms to detect animal sounds. For example, lions have eight or nine words they use to describe different behaviours. Our student developed a lion call to English translator so people can understand what lions are saying. In other projects they have developed models to count animals using computer vision to estimate abundance. One student also worked on a project about interspecies money where orangutans have their own wallet which they use to pay local guardians for the care giving services they provide. As long as the orangutan stays alive and well and its habit is maintained, then local communities will receive money. The student developed a costing model based on computer vision.
What are the course's unique selling points?
The course is applied and it exposes students to highly sought-after industry related tools and techniques. We are both NVIDIA-certified instructors and university ambassadors for the NVIDIA Deep Learning Institute, which is world renowned for delivering cutting edge deep learning tools and training. This allows us to issue NVIDIA certification in deep learning to our students alongside their MSc degree. All the techniques provided by NVIDIA in deep learning are taught to our students, such as RAPIDS, Triton Server and LLM model hosting and inference.