From: mk_thisisit
The emaza application is an open-source tool developed by Poles to aid in the preservation of endangered species, primarily in the forests of Central Africa [00:00:00], [00:00:49]. It has been recognized as one of the top 10 applications globally by UNESCO [00:00:40], [00:00:52].
Empaza: Overview and Purpose
emaza was created by the company Apsilon in collaboration with scientists from the University of Stirling and national parks in Gabon [00:01:16]. Its primary goal is to empower scientists and park guards to study deep tropical forests using modern technologies, specifically machine vision and machine learning [00:01:32], [00:01:39].
The application serves several key purposes:
- Biodiversity Preservation It helps in activities aimed at preserving biodiversity, such as saving endangered species [00:05:44], [00:05:46].
- Tracking Animal Migrations By identifying animals and their locations, it allows researchers to understand animal migration patterns, such as those of elephants, which are being impacted by climate change [00:05:07], [00:05:12], [00:05:19]. This information can prevent the construction of roads in critical migration routes [00:05:51].
- Disease Monitoring It assists in studying the spread of diseases within animal communities, such as gorilla populations. The application quickly identifies gorillas in photos, allowing experts to assess their health and understand disease progression [00:05:59], [00:06:09].
How Empaza Works
The application streamlines the process of analyzing data collected from camera traps placed in vast rainforests.
Traditional Manual Process
- Deployment: Ecologists and forest guards embark on expeditions, often lasting 3-4 weeks, to set up small, palm-sized camera traps throughout the forest. These traps are equipped with cameras, heat sensors, and motion sensors that trigger a photo when an animal, such as an African forest elephant, passes by [00:02:07], [00:02:20], [00:02:22], [00:02:25].
- Data Retrieval: After a few weeks (typically 3-4), the teams retrace their steps to collect the SD cards from the camera traps [00:02:41], [00:03:02].
- Manual Analysis: Back at their forest base, often without internet access, scientists manually review thousands of photos. A single expedition can yield around 30,000 photos [00:03:10], [00:03:19]. This process typically takes 2-3 weeks [00:03:35]. Manual review is slow and error-prone due to the sheer volume, visual complexity of forest environments (bushes, branches, sun spots), and human fatigue, making it difficult to spot animals, especially smaller ones [00:03:46], [00:03:49], [00:03:59].
Empaza’s Automated Process
emaza leverages machine learning algorithms, specifically trained to recognize animal species in photos [00:01:41], [00:01:43].
- Offline Processing: The application runs completely offline on a standard laptop in the forest base [00:04:14], [00:04:17].
- Rapid Analysis: Users select the appropriate model and let the computer process the photos. A batch of 30,000 photos, which would manually take 15 days, can be processed by
emazain approximately 8 hours, or about one day [00:04:36], [00:04:39], [00:04:42], [00:04:44]. - Accuracy: While machine learning models are not perfect, their accuracy is comparable to human accuracy, and any errors tend to balance out. Ecological models built using
emaza’s output are of comparable quality to those derived from manual analysis, with the added benefit of significant time savings [00:08:59], [00:09:01], [00:09:07], [00:09:11], [00:09:17].
Global Reach and Open Source Nature
The emaza project’s full source code, including its machine learning models, is openly available on GitHub [00:06:27], [00:06:32], [00:06:35]. This open-source approach aims to maximize its positive impact on nature conservation by allowing the international community, scientists, and park guards to use and further develop the application [00:06:42], [00:06:45], [00:06:51], [00:10:06].
Current Usage
- Africa: Primarily used in Gabon, Cameroon, and Kenya, and recently tested in Congo [00:00:22], [00:06:57].
- Antarctica: A project in cooperation with the Arctowski station of the Polish Academy of Sciences focuses on detecting animals from drone images [00:00:30], [00:00:35], [00:13:01], [00:13:04], [00:13:06].
- North Pole: Research is conducted to accelerate the study of plankton on a large scale, vital for ocean life and climate [00:13:21], [00:13:25].
UNESCO Recognition
In 2023, the International Institute for Research on Artificial Intelligence, under the auspices of UNESCO, selected emaza as one of the “outstanding” projects globally. This places it among the top 10 out of 100 distinguished projects that significantly contribute to the UN Sustainable Development Goals [00:07:10], [00:07:14], [00:07:21], [00:07:24]. emaza specifically supports two of these goals: preservation of biodiversity and the protection of life on land [00:07:34], [00:07:39], [00:07:41].
Origin and Funding
The idea for emaza originated when a post-doctoral ecologist from the University of Stirling in Scotland, specializing in tropical forest research in Gabon, approached Apsilon. He recognized the potential for automatic data processing from camera traps but lacked the technical expertise [00:08:08], [00:08:15], [00:08:20], [00:08:26]. Apsilon, a Polish company, developed machine learning models that proved capable of automating the process with high accuracy, comparable to manual methods [00:08:48], [00:08:50], [00:09:01].
The application was created as part of Apsilon’s “Data for Good” program, a pro bono initiative [00:10:13], [00:10:15], [00:11:41]. While there has been some funding from non-governmental and governmental institutions, this has only covered small fragments of the development, making it largely a pro bono effort [00:11:54], [00:12:00], [00:12:02]. Apsilon’s main income is derived from commercial projects in data visualization (using R Shiny technology) and other machine learning applications, particularly in computer vision and image analysis [00:10:22], [00:10:51], [00:11:02], [00:11:28], [00:11:33], [00:11:37].
Role of AI in Sustainable Development
Artificial intelligence plays a significant role in addressing sustainable development goals, extending beyond business applications. Many projects leverage machine learning to tackle environmental challenges, particularly where large datasets require rapid analysis [00:12:06], [00:12:34], [00:12:36], [00:12:51]. These applications enable ecologists and other scientists to gain critical insights into complex natural systems that would otherwise be impossible to observe, such as animal movements or the health of plankton, which has a huge impact on climate [00:05:28], [00:05:30], [00:13:27], [00:13:30].