About This Project
The Desert Lion Conservation Project uses camera traps to monitor, protect and maintain the welfare of lions living in the Northern Namib Desert. Since analysing camera trap images demands a substantial amount of effort, Smart Parks was asked to develop a software tool that can automatically detect lions with minimal human interference. The resulting software will be published open-source.
Ask the Scientists
Join The DiscussionWhat is the context of this research?
Nature conservationists all over the world use camera traps to monitor wildlife populations. Since analysing this data often demands a substantial amount of effort, computer vision models are becoming increasingly popular to automate this process. For many non-coding conservationists, however, it has proven difficult to implement these models since most of the interaction is exclusively through Python code and command-line interfaces. The current project will develop an open-source platform to overcome this limitation - with a pilot study for the Desert Lion Conservation Project in Namibia.
What is the significance of this project?
Smart Parks was asked to ensure the development and delivery of (i) an accurate baseline machine learning model that automatically detects lions from camera trap images in the unique environment of the Namib Desert, and (ii) a non-coding platform to keep improving this baseline model with new data over time. The principal investigator created EcoAssist, creating a free software package where researchers can create their own species detection models. EcoAssist represents an important step forward in conservation technology. Its collaborative approach, transferability to other species and habitats, and leveraging of our organisation's expertise make it a promising tool for supporting wildlife conservation efforts globally.
What are the goals of the project?
The funds will be used to purchase a dedicated machine learning computer capable to train and deploy computer vision models. The project is currently in its orientation phase and will start with its model training phase soon. The outcome of this project is not only a computer vision model capable of detecting species in Northern Namibia, but also fine-tuning and documenting the process, and further development of the open-source AI-platform EcoAssist. Through EcoAssist, ecologists all over the world can train and deploy their own models.
Budget
In order to automatically detect lions (Panthera leo) from camera trap images in the unique environment of the Namib Desert, a computer vision model will need to be trained on several tens of thousands of images. By acquiring a dedicated machine learning computer, we can process these images. This machine will be used to analyse any future images and retrain models with new data. At the moment, the goal is to develop a model which can accurately detect lions. However, there are aspirations to enlarge the model with more species (hyena, leopard, giraffe, elephant, etc.) to develop an automated monitoring tool for the Desert Lion Conservation. This computer will be used for the entire process.
Endorsed by
Project Timeline
This project effectively began in January 2022, when I published the first version of EcoAssist. However, only since I partnered with Smart Parks and Desert Lion Conservation in July 2023, there is the plan to actively train species detectors for existing nature conservation organisations. The model will be implemented in the existing data flow of the Desert Lion Conservation project. Throughout the project, the software will be updated on GitHub in real time.
Jan 12, 2022
Create EcoAssist
Jun 01, 2023
Partner with Smart Parks and Desert Lion Conservation
Oct 01, 2023
Orientation phase
Oct 03, 2023
Project Launched
Nov 01, 2023
Training phase
Meet the Team
Team Bio
Smart Parks provides advanced sensor solutions to conserve endangered wildlife and efficiently manage park areas by providing cutting-edge technologies.
Peter van Lunteren
My name is Peter and I'm a tech-savvy ecologist with a passion for data science and machine learning. I'm dedicated to developing software that aid nature conservationists in the field. Currently, I'm working on EcoAssist, an AI platform that enables annotation, training, and deployment of custom computer vision models for automatic species detection, offering ecologists a way to save time reviewing images and focus on conservation efforts. At the time of writing, approximately 4500 ecologists around the world already use EcoAssist in their workflow. My goal is to advance EcoAssist to a well-known toolkit for conservation projects around the world.
Lab Notes
Nothing posted yet.
Additional Information
In the field of species detection models for wildlife conservation, there are other existing solutions, such as TrapTagger and WildMe, which also offer valuable contributions. However, EcoAssist sets itself apart through several key differentiators. While both TrapTagger and WildMe provide species detection models, they require images to be uploaded to the cloud for analysis. In contrast, EcoAssist is designed to be completely offline and runs on the local machine. This feature makes EcoAssist ideal for organisations and park managers that operate in remote areas with limited or unstable internet connectivity, where cloud-based solutions may not be feasible.
Another standout feature of EcoAssist is its user-centric approach. Unlike generic models provided by other platforms, EcoAssist empowers users to train their own models. This feature enables researchers and conservationists to fine-tune the detection algorithms precisely for their specific project needs and target species. This flexibility makes EcoAssist adaptable to the needs of different conservation efforts.
Project Backers
- 2Backers
- 1%Funded
- $20Total Donations
- $10.00Average Donation