Contents:

Free Python Course ➞ Mini-course for beginners and experienced coders. 4 cool projects in the portfolio, live communication with the speaker. Click and find out what you can learn in the course.
Learn MoreData science, known as Data Science, provides modern professionals with powerful tools to combat global threats. In this field, data analysts and programmers play a key role in solving environmental problems and protecting the environment. They apply big data analytics to monitor the health of ecosystems, predict climate change, and assess the impact of human activity on nature. Using machine learning algorithms and data visualization, they help identify trends and develop strategies for sustainable development. Thus, Data Science is becoming an important ally in efforts to protect our planet.
Protect rare animals
The snow leopard, also known as the irbis, is one of the most beautiful animals on the planet. These elegant predators are endangered, with fewer than 4,000 individuals currently living in the remote mountainous regions of Central Asia. Studying snow leopards is a real challenge, as these animals are masters of camouflage. Their stealth has earned them the nickname "ghosts of the mountains." Conservation of the snow leopard has become an important environmental issue that requires attention and protection of its natural habitat.

In Zoologists have installed numerous automated cameras with motion sensors in snow leopard habitats. However, there is a problem: the cameras record any activity, such as flying birds. As a result, approximately 500,000 images are accumulated annually, only 5% of which contain images of snow leopards. Scientists must spend hundreds of hours reviewing all the photographs and selecting the most useful ones. Optimizing the data processing process can significantly reduce the time and increase the efficiency of snow leopard research.
To automate the process of snow leopard monitoring, Microsoft, in collaboration with the Snow Leopard Trust, a foundation for the protection of these animals, has developed a system that ensures 95% accuracy in detecting snow leopards in photographs. This innovative technology allows scientists to reduce the time spent on routine image processing and focus on studying the population and behavior of this rare cat species. As a result, researchers are able to more deeply analyze the ecosystem and develop effective strategies for conserving snow leopards in their natural habitat.
In 2020, Russia implemented a similar system to study the Far Eastern, or Amur, leopard—a rare species of big cat found on the country's border with China and North Korea. This initiative aims to preserve and study the endangered Amur leopard population. The system allows for more effective tracking of the migration and behavior of these animals, which in turn facilitates the development of measures to protect them and preserve their natural habitat.
In the computer vision system, Microsoft specialists used convolutional neural networks (CNN), which have proven effective in image processing tasks. Despite these advances, research into optimal CNN use continues, opening new horizons for improving the accuracy and speed of visual information processing.
Transfer learning is actively used in training neural networks in data analytics. This approach allows a model to be trained to solve one problem and then adapted to solve another, transferring knowledge from one area to another. This method allows the neural network to achieve high performance on a new task with minimal additional training. Transfer learning significantly accelerates the development process and improves model quality, making it an essential tool in modern machine learning.
The programmers used the ResNet neural network architecture, first training it on images of various objects unrelated to snow leopards. They then further trained the model using images of snow leopards. During the further training process, some of the images were flipped, allowing the network to recognize snow leopards in any orientation. Using the Transfer Learning method resulted in a 20% improvement in the neural network's performance.
To improve the efficiency of data analysis, data science experts recommend treating images taken at the same time as a single set of images. This methodology is based on the assumption that if a snow leopard is captured in one image, the likelihood of its appearance in subsequent images increases significantly. This approach allows for more accurate tracking of the behavior and movements of rare animal species, which in turn contributes to a deeper understanding of the ecosystem and conservation methods.
Neural networks can make errors when analyzing images, for example, failing to recognize a snow leopard in a single photograph. However, when analyzing a group of images as a whole, the probability of detecting an object increases significantly. This approach increases the accuracy of the convolutional neural network to 95%, virtually eliminating false positives. This underscores the importance of comprehensive image analysis for achieving high-performance object recognition.
Zoologists can now track the movements of snow leopards in real time. When a snow leopard is detected in camera images, the information is displayed on an interactive map, indicating the exact location and time of capture. The efficiency of this process is ensured by cloud technologies that connect all system components. This innovation allows for better study of the behavior of these rare animals and the development of conservation measures.
In the near future, scientists plan to train a neural network to recognize snow leopards based on their unique body characteristics and spot patterns. This project aims to improve snow leopard population monitoring and conservation. The use of neural network technologies will allow for more accurate identification of individual animals, which is important for ecology and wildlife conservation research.
Catch the Poachers
About 180 million tons of fish are caught worldwide each year. International agreements and national laws exist to regulate fisheries and limit catches. However, enforcing these standards on the high seas is a significant challenge. Difficulties in monitoring fisheries create conditions for illegal fishing, which negatively impacts marine ecosystems and bioresources. Sustainable fisheries management requires more effective monitoring methods and international cooperation to conserve marine resources.
Poaching vessels, forming so-called "dark fleets," actively operate in known fishing grounds. Illegal fishing threatens to upset the delicate balance of the marine ecosystem, which can lead to devastating consequences for biodiversity and the sustainability of marine resources. Protecting the marine environment and combating poaching are becoming priorities for preserving the health of oceans and fishing grounds.
In 2016, Google launched Global Fishing Watch, a service that tracks fishing vessels across the world's oceans in real time. This service displays both legal and illegal vessels, providing the ability to track their location and fishing method at any time. Vessel type and activity are determined using artificial neural networks, ensuring high data accuracy and transparency in the fisheries sector. Global Fishing Watch plays a vital role in combating illegal fishing and promoting marine conservation.

The creation of the Global Fishing Watch service required the integration of various technologies, including satellite tracking, geolocation, cloud computing, machine learning, data visualization, geographic information systems, and web technologies. These technologies enable effective monitoring of fisheries, analysis of marine resource data, and increased transparency in this area. The use of modern solutions ensures the accuracy and reliability of information, which contributes to the sustainable management of fish stocks and the protection of the marine ecosystem.
The service relies on vessel location data obtained from multiple sources. The majority of the information (80%) comes from Automatic Identification System (AIS) signals. AIS allows one to determine a vessel's coordinates, course, and speed. Most large vessels are equipped with AIS transmitters, which ensure safe navigation by preventing collisions at sea. Every day, the service processes 60 million messages received from 300,000 vessels, which helps provide up-to-date and accurate data on maritime transport movements.
National vessel movement monitoring systems are an important source of information. Many countries are actively connecting to this service and transmitting data from their tracking systems. Satellites used by Google for Earth mapping are used to locate vessels without identification systems or with their systems disabled. These satellites are equipped with infrared, radar, and optical technology, allowing them to detect objects at sea both at night and in dense cloud cover. The use of such technologies contributes to improved maritime safety and the efficiency of maritime transport management.
To efficiently store and process large volumes of data, data analysts use Google cloud services. Data is visualized on a map using the company's geoservices, allowing for clear presentation of information and simplifying its analysis. Google cloud technologies provide high performance and reliability, making them an ideal choice for working with large data sets.
Simply marking a point on a map is not enough. It is important to identify the type of vessel, understand its fishing methods, and determine the moments when vessels enter prohibited areas to begin fishing. Performing this task manually is extremely difficult due to the enormous volume of data. For effective analysis, it is necessary to use modern technologies and algorithms that will help automate the process and improve the accuracy of identifying fishing violations.
To automate the process, data science specialists developed a training dataset in which the moments of the start of fishing were marked on the vessel trajectories. At this moment, the vessel's trajectory changes, which distinguishes it from the usual movement between ports. By analyzing the trajectories, it is possible to determine the fishing method used by the vessel. The project team manually classified several thousand ship trajectories, which formed the basis for further automation and increased accuracy in identifying fishing methods.
Data analysts trained a neural network using the labeled data, creating a system capable of identifying vessel types, such as cargo ships, trawlers, longliners, and others. It also records the start and end times of marine animal harvests. This neural network demonstrates high accuracy, reaching over 90% in its decisions, which is comparable to the work of an experienced expert in this field.
The system effectively counters attempts by some fishermen to conceal their activities. Fishermen often change their home ports, vessel names, and use false AIS numbers. However, thanks to machine learning technologies, Global Fishing Watch can identify unique vessel movement characteristics, making them virtually impossible to evade control.
The service is constantly being updated and improved. In the near future, the system will learn to detect more types of suspicious activity. One such activity is the illegal transfer of caught fish from one vessel to another on the high seas. This innovation will help improve the effectiveness of enforcement of fishing laws.

