From: mk_thisisit
The application “emaza” is a unique project created by Poles to assist in preserving biodiversity, primarily focusing on saving endangered species in the forests of Central Africa [00:00:00], [00:01:05]. It was recognized by UNESCO as one of the top 10 applications globally for its contribution to sustainable development [00:00:40], [00:00:52].
How emaza Works
Developed by the Apsilon company in cooperation with scientists from the University of Stirling and national parks in Gabon, emaza allows scientists and park guards to study tropical forests using modern technologies like machine vision [01:16:00], [01:34:00].
Camera Trap System
Ecologists and forest guards place camera traps—small devices with cameras, heat sensors, and motion sensors—deep within forests [02:20:00]. These traps capture photos when an animal, such as an African forest elephant, passes by [02:30:00]. After several weeks, scientists return to collect the SD cards containing tens of thousands of photos [02:44:00], [03:02:00], [03:19:00].
Machine Learning for Data Processing
Manually reviewing 30,000 photos from a single expedition typically takes two to three weeks, as it is a tiring and attention-demanding task due to the complex forest backgrounds [03:35:00], [03:46:00], [04:09:00]. emaza utilizes machine learning algorithms trained to recognize species in these photos [01:39:00], [01:43:00]. The application processes 30,000 photos in about 8 hours (one day) while operating completely offline, significantly reducing processing time from 15 days to one [04:14:00], [04:36:00], [04:44:00]. This efficiency allows scientists to quickly move to extracting valuable information [04:49:00].
Impact on Conservation
The accelerated data processing by emaza provides crucial insights for conservation efforts:
- Mapping Animal Movements: By identifying which camera trap captured which animal and when, scientists can map animal appearances on a geographical map to understand migration patterns, such as those of elephants, which are changing due to climate change [04:55:00], [05:07:00], [05:16:00]. This knowledge, otherwise unattainable, helps prevent building roads in critical migration routes [05:27:00], [05:48:00].
- Disease Monitoring: The application helped identify photos of gorillas during a disease outbreak in their community, enabling experts to quickly assess their health and understand the disease’s spread to implement aid [05:59:00], [06:06:00], [06:15:00].
Global Reach and Open Source
Emaza is primarily used in Gabon, Cameroon, and Kenya, and has recently been tested in Congo [00:21:00], [06:56:00]. Beyond Central Africa, it is also being used in cooperation with the Arctowski station in Antarctica for detecting animals from drone images [00:28:00], [01:03:00], and research at the North Pole focuses on accelerating the study of plankton [01:31:00], [01:31:00].
The application is open source, with its full source code and machine learning models available on GitHub [06:27:00], [06:32:00], [06:35:00]. This ensures the international community can develop it further and maximize its positive impact on nature conservation [06:42:00], [06:49:00].
UNESCO Recognition
Emaza was distinguished by the International Institute for Research on Artificial Intelligence (under UNESCO auspices) as one of the “outstanding” projects globally. This places it in the top 10 among 100 distinguished projects that contribute most to the UN Sustainable Development Goals [07:06:00], [07:21:00], [07:57:00], [08:01:00]. Emaza specifically aids in achieving goals related to the preservation of biodiversity and the protection of life on land [07:34:00], [07:39:00].
Development and Accuracy
The idea for emaza emerged when an ecologist from the University of Stirling, researching tropical forests in Gabon, approached the Apsilon company with the concept of automatically processing camera trap data [08:08:00], [08:15:00], [08:26:00], [08:30:00]. Apsilon’s expertise in machine learning allowed them to develop models that achieved accuracy comparable to manual human classification [08:48:00], [08:59:00]. Although machine learning models can make mistakes, these errors tend to balance out, leading to ecological models built on this information being at a comparable scientific level [09:11:00], [09:17:00], [09:22:00]. A paper detailing these findings was published in the prestigious scientific journal Ecology and Evolution [09:25:00].
Company Model
Apsilon’s “Data for Gold” program supports pro-bono projects like emaza [01:15:00]. The company primarily specializes in building decision support systems (interactive dashboards for complex data visualization) and commercial projects related to computer vision and image analysis [01:22:00], [01:31:00], [01:33:00], [01:37:00]. While some small fragments of emaza’s development have received funding from non-governmental and governmental institutions, it is largely a pro-bono effort [01:50:00].