Detecting Wildfires Using CNN Model with 95% Accuracy

Detecting Wildfires Using CNN Model with 95% Accuracy

How wildfires detection company Sintecsys leveraged Omdena’s community to build a fire detection algorithm in two months using AI and a CNN Model.

 

 

The Problem: Wildfires and a Convolutional Neural Network

2019 was marked by very big fires. Not only the Notredame cathedral in Paris, and the National Museum in my country Brazil but entire complex ecosystems like the Amazon forest Wildfires and more recently in Australia. Before we dive into our finished product of, how to detect and stop wildfires early on with our community-build AI tool, let us understand how forest fires start.

  • Natural fires: Generally, natural fires are started by lightning, with a small portion originated by spontaneous combustion.
  • Human-caused fires: Humans cause fires in multiple ways such as smoking, recreation, soil preparation for agriculture, and so on. Man-caused fires represent the greatest percentual share of fires, but natural-caused fires represent larger burned land areas. This happens because the man-caused are detected earlier, while natural fires can take hours to be identified by the competent authorities.

Regardless of the causes, when a forest like in the Amazon starts to burn, the fire can spread and reach speeds of up to 23 km/h and reach temperatures of 800 °C (1470 °F) destroying plant and animal life within a few hours (sometimes even contributing to species extinction)

Even worse, fires damage the planet through CO2 that will contribute to global warming.

In addition to disrupting the climate, it impacts the sky and the quality of the air of a huge metropolitan city like São Paulo, the most important economical and productive center for my country.

At 3 pm, August 19th, 2019, a black sky appeared as a result of the meeting of a cold front with the fire particulates stemming from the Amazon and midwest fires in my country.

The day became night, and the feeling was that we were living in a biblical plague as described in the Old Testament. Really scary!

 
 
 

 

Sao Paulo's sky warning for wildfires

Pictures of São Paulo’s sky at 3 pm on August 19th, 2019

 

Among much misinformation, one post from NASA stood out by shedding the fundamental light of science on the matter.

In the image below, you see a colored high-resolution satellite image showing how the fire smokes spread to the southeast states of my country.

 

VIIRS image given by NASA

By NOAA/NASA’s Suomi NPP using the VIIRS (Visible Infrared Imaging Radiometer Suite) on August 20th, 2019

 

The Solution

Sintecsys´s growing customer base of clients on farms and forests can confirm. The company installs cameras on top of communication towers to capture images that are sent to a monitoring center. Once there is fire (or smoke) detected on images, it sends alerts and fire fighting actions. This saves lives and infrastructure costs.

Sintecsys is not alone in its mission as there are many other companies around the world dedicated to this mission also in a very successful way.

The company installed 50 towers distributed in Brazil (2019 data.

To extend the customer reach and scale their business model to thousands of cameras with the capability of accurately and quickly detecting wildfire outbreaks, Omdena’s AI capabilities come into play.

Omdena is a global platform where organizations collaborate with a diverse AI community to build solutions for real problems in a faster and more effective way.

 

 

#1 Scoping the problem

To tackle this problem, Omdena and Sintecsys agreed to deal with day images in their first joint challenge and in a second challenge improve the solution by dealing with night images.

The main difference between day and night images for fire detection is that during the day images usually show smoke and during the night these images show live fire. Both sunset and dawn, where smoke and live fire coexist on images, represent boundary conditions for the problem.

#2 Working on the dataset

The dataset was really big comprising footage and images from different cameras with and without fires outbreaks happening. Combining the original images given, our team had almost 7.600 images of 1920 x 1080 size (day images without fires outbreaks, day images with fires, and some night images (around 16%)) to start labeling.

 

 

Data set samples from sao paulo

Samples from the datasets

 

 

To add even more images, Gary Diana built an algorithm to successfully extract images from the footage and at the same time avoiding the generation of images with the same landscape among them (de-duplication). This initiative added another 1.150 images of 1280 x 720 size to our dataset.

#3 Labeling with Labelbox

Having the datasets prepared for labeling, we gathered around 20 people dedicated to the task, created the environments on Labelbox, which is the best tool available for computer vision by allowing labeling data, managing its quality and operating a production training data pipeline, and then, at last, we started to make tests and to label the final datasets.

I managed the task but I received huge support from Alyona Galyeva who helped the whole team not only by labeling but also by reviewing and managing everyone´s work.

In her own words:

It always starts with a mess when a group of people collaborates on a labeling project. In our case, Labelbox saved us a lot of time and effort by not allowing multiple users to label the same data. On top of that, it made our lives easier by proposing 4 roles: Labeler, Reviewer, Team Manager, and Admin. So, nobody was able to mess with data sources, data formats, and, of course, the labels made by other people.

 
 
 

Labelbox interface for labeling, managing and reviewing labels

 

Having both datasets labeled, the next train, validation, and test files were generated by the data pipeline team.

#4 Building the models

From the start, the team searched and studied several top-notch papers with different techniques that could be applied to solving the problem.

The challenge team created several teams in different tasks, each one focused on trying different approaches: mobile net, semantic segmentation, Convolutional Neural Networks (CNNs) — from simple architectures to more sophisticated ones.

Another great testimony of this step comes from Danielle Paes Barretto:

It was inspiring to see people eager to achieve great results. I tried to help in all tasks; from labeling the data to building CNN models and testing them on our dataset. We also had frequent discussions which in my opinion is one of the greatest ways of learning. All in all, it was an amazing opportunity to learn and to use my knowledge for the good while meeting great people!

In addition, different techniques were successfully applied to improve results like creating patches of different sizes on original images and training over patches, data augmentation (e.g. horizontal and vertical flipping), denoising images, etc.

#5 Results

The final solutions were able to reach a recall between 95% and 97% while having a false positive rate between 20% and 33%, which means that these solutions were extremely successful in catching 95% to 97% of the real fires outbreaks. While the challenge partner Sintecsys is extremely happy with the results, in our second challenge, we will improve the current models by adding night time images.

More about Omdena

Omdena is an innovation platform for building AI solutions to real-world problems through the power of bottom-up collaboration.

Omdena Delivers Technology to Detect Wildfires in the Amazon Forest

Omdena Delivers Technology to Detect Wildfires in the Amazon Forest

Leader in Collaborative AI Solutions Omdena Delivers Technology to Detect Wildfires in the Amazon Forest.

Omdena Logo

 

By Laura Clark Murray 


 

 

For Immediate Release

February 26, 2020

Leader in Collaborative AI Solutions Omdena Delivers Technology to Detect Wildfires in the Amazon Forest.

Omdena delivers eleventh completed project challenge, and the second related to fire-detection, since May 2019 launch Palo Alto, California — Omdena, the global platform for building collaborative AI-based solutions, today announced the delivery of AI technology to detect the outbreak of wildfires in Brazil. Created in partnership with Sintecsys, a commercial agriculture technology company monitoring 8.7M acres across four Brazilian biomes, the system identifies flames and smoke in images with more than 95% accuracy.

Omdena’s diverse team of 47 data scientists from 22 countries joined Sintecsys’ small internal AI group for the eight-week machine learning project. With this challenge, Omdena builds on an impressive record of building artificial intelligence solutions through global collaboration. To date, more than 800 people from more than 75 countries have participated.

“Omdena provided Sintecsys with a ten-fold increase in the size of their AI team,” said Rudradeb Mitra, Founder of Omdena. “Our diverse experts brought deep AI knowledge and experience to solve the problem of accurately detecting wildfires from real-time images.”

The longer a wildfire burns, the more damage it can cause and the more difficult it is to contain. While Sintecsys had dramatically reduced detection time from 40 minutes to under 5, a high incidence of false positives meant a delay in calling in firefighters until a fire could be confirmed. The accuracy of the fire-detection system built with Omdena will allow Sintecsys to alert firefighters within the first few minutes of an outbreak.

“Every second counts in order to preserve life,” said Osmar Bambini, Sintecsys Head of Innovation. “Speed, accuracy, and power sum up my perception of Omdena. For Sintecsys, from now on Omdena is the official AI partner.”

###

For media inquiries contact: Laura Clark Murray, Omdena, laura@omdena.com

About Omdena: Omdena is a platform for building AI solutions to real-world problems through global collaboration. Our partners include the UN World Food Programme and the UN Refugee Agency. Omdena is an Innovation Partner of the United Nations AI for Good Global Summit 2020. Learn more at Omdena.com.

Learn more about Omdena’s Wildfire Detection AI Challenge at https://omdena.com/blog/wildfires-artificial-intelligence 

About Sintecsys: Sintecsys is a commercial agri-tech firm delivering a monitoring system that detects fire outbreaks quickly and effectively. Learn more at https://sintecsys.com/en/home-en/

Omdena Delivers Technology to Detect Wildfires in the Amazon Forest

Detecting Wildfires with Artificial Intelligence

How a small team leveraged Omdena’s AI community to detect wildfires in the Amazon forest.

By Laura Clark Murray

 

What can be done about wildfires?

 

The power of wildfires to destroy has been tragically apparent around the globe. Depending on terrain and conditions, they can double in size every 10 minutes. And every minute that a fire burns makes it harder to contain. With early detection and a quick response, a fire may be quickly extinguished. In contrast, a fire left unchecked can wipe out huge amounts of forest, killing and destroying as it rapidly spreads.

 

How do you stop a fire before it becomes wild?

It’s the job of Sintecsys, a commercial agriculture technology company in Brazil, to monitor 8.7 million acres of forest and agricultural land across four biomes, including the Amazon forest. In order to identify fire, their system works around the clock to process images from 360-degree cameras mounted on towers distributed throughout that land. If there appears to be flames or smoke, the system alerts the staff. As a result, in the last 3 years they’ve dramatically reduced fire detection time from an average of 40 minutes to under 5. 

 

Sample images Sintecsys

 

Osmar Bambini, Head of Innovation at Sintecsys, knew that artificial intelligence could be used to reduce that detection time further. In addition, AI held promise in separating genuine cause for alarm from false alarms. In order to avoid missing any actual fires, the system triggered a high rate of “false positives”. Therefore the staff needed to validate each fire alert, before calling in firefighters. This extra processing delayed the response to real fires by valuable minutes.

 

How can a company apply AI without a large in-house team?

Bambini brought on Omdena. Omdena is a global platform where AI experts and data scientists from diverse backgrounds collaborate to build AI-based solutions to real-world problems. You can learn more here about Omdena’s innovative approach to building AI solutions through global collaboration. 

For this eight-week machine learning project, Omdena pulled together a diverse team of 47 data scientists from 22 countries to join Sintecsys’ small internal AI group. Notably, a data scientist in Brazil, Leonardo Sanchez, was eager to join the Omdena challenge. It gave him the opportunity to address a problem of significance for his country and the world. You can read about his perspective, and the image processing approaches behind the Omdena solution, in his article “How to Stop Wildfires with Artificial Intelligence”.

Yash Mahesh Bangera and Ashish Gupta had their own reasons for becoming Omdena collaborators. Specifically, they have a dream of working with organizations that undertake initiatives for social good. Joining this Omdena challenge allowed them to do just that. Moreover, they deepened their own AI and machine learning skills in the process, as they explain in their article on the project.

In two months time, the team built a system that is accurate in identifying smoke and flames in daytime images more than 95% of the time. Due to that accuracy, false positives are dramatically reduced. As a result, firefighters can be called onto the scene without delay. Bambini is thrilled with the results: “Outstanding! The Omdena challenge provided the Sintecsys team an intense and accurate deep dive into AI with amazing results.” By early March, the AI system will be fully deployed.

 

What’s next?

Sintecsys and Omdena are exploring a second project which will tackle the detection of smoke and fire outbreak in nighttime images. In addition, we’ll pull in satellite imagery to get a more complete view of what’s happening on the ground.

Bambini has big plans for making the system even smarter with follow-on projects with Omdena. We’ll use AI to identify areas in the forest that are especially high-risk for a fire. Above all, human activity is the most significant indicator of fire risk. “More than 90% of fires are caused by humans, either intentionally or accidentally. The places where farmers have been clearing land and where people are settling are the highest risk spots,” says Bambini. “If we can use AI to pinpoint those areas, we’ll be able to predict where fires are most likely to happen.” 

If Bambini is looking in the right place, he just might be able to detect a fire the moment it breaks out.

“Speed, accuracy, and power sum up my perception of Omdena”, says Osmar Bambini, Sintecsys Head of Innovation. “For Sintecsys, from now on Omdena is the official AI partner.”

 

Learn More

Keep up with our work with Sintecsys to refine their fire detection system with AI hereThis is Omdena’s second fire-related challenge. Read about our work with Swedish AI startup Spacept to prevent fires sparked by falling trees near power lines.


Want to work with us? Tell us about your project here.
The video sums up the project. You can find it on LinkedIn here.

 


About Omdena

Omdena is a collaborative platform where organizations work with a diverse AI community to build solutions for real-world problems.

Learn more about us and Collaborative AI.

AI Applied: How-to Stop Wildfires through Flame and Smoke Detection

AI Applied: How-to Stop Wildfires through Flame and Smoke Detection

How Brazilian company Sintecsys worked with Omdena’s AI community to build a fire detection algorithm to stop wildfires before an outbreak can occur.

Article written by Leonardo Sanchez

 

 

 

 

The Problem

2019 was marked by very big fires.

Not only the Notredame cathedral in Paris, and the National Museum in my country Brazil but entire complex ecosystems like the Amazon forest fires and more recently in Australia.

Before we dive into how to detect and stop wildfires early on with our community-built AI tool, let us understand how forest fires start.

  • Natural fires: Generally, natural fires are started by lightning, with a small portion originated by spontaneous combustion.
  • Human-caused fires: Humans cause fires in multiple ways such as smoking, recreation, soil preparation for agriculture, and so on. Man-caused fires represent the greatest percentual share of fires, but natural-caused fires represent larger burned land areas. This happens because the man-caused are detected earlier, while natural fires can take hours to be identified by the competent authorities.

Regardless of the causes, when a forest like in the Amazon starts to burn, the fire can spread and reach speeds of up to 23 km/h and reach temperatures of 800 °C (1470 °F) destroying plant and animal life within a few hours (sometimes even contributing to species extinction).

Even worse, fires damage the planet through CO2 that will contribute to global warming.

In addition to disrupting the climate, it impacts the sky and the quality of the air of a huge metropolitan city like São Paulo, the most important economical and productive center for my country.

At 3 pm, August 19th, 2019, a black sky appeared as a result of the meeting of a cold front with the fire particulates stemming from the Amazon and midwest fires in my country.

The day became night, and the feeling was that we were living in a biblical plague as described in the Old Testament. Really scary!

Pictures of São Paulo’s sky at 3 pm on August 19th, 2019

 

Among much misinformation, one post from NASA stood out by shedding the fundamental light of science on the matter.

In the image below, you see a colored high-resolution satellite image showing how the fire smokes spread to the southeast states of my country.

By NOAA/NASA’s Suomi NPP using the VIIRS (Visible Infrared Imaging Radiometer Suite) on August 20th, 2019

 

In Brazil and many other places in the world, we have seen that fires left thousands of homeless, resulted in many deaths, property damage and unfortunately, it will not be the last time in human history for devastating fires to spread.

 

How can AI help to stop wildfires?

Is it possible to help my country (and other countries)? Is there a way to use the power of community and of AI to achieve this?

 

The project

According to the Brazilian wildfire detection company Sintecsys: yes!

Sintecsys´s growing customer base of clients on farms and forests can confirm. The company installs cameras on top of communication towers to capture images that are sent to a monitoring center. Once there is fire (or smoke) detected on images, it sends alerts and fire fighting actions. This saves lives and infrastructure costs.

Sintecsys is not alone in its mission to stop wildfires as there are many other companies around the world dedicated.

So far, the company installed 50 towers distributed in Brazil (2019 data).

To extend the customer reach and scale their business model to thousands of cameras with the capability of accurately and quickly detecting wildfire outbreaks, Omdena’s AI capabilities come into play.

Omdena is a global platform where organizations collaborate with a diverse AI community to build solutions for real problems in a faster and more effective way.

 

How the team solved the problem

#1 Scoping the problem

To stop wildfires early on before further damage is caused, Omdena and Sintecsys agreed to deal with day images in their first joint challenge and in a second challenge improve the solution by dealing with night images.

The main difference between day and night images for fire detection is that during the day images usually show smoke and during the night these images show live fire. Both sunset and dawn, where smoke and live fire coexist on images, represent boundary conditions for the problem.

 

#2 Working on the dataset

The dataset was really big comprising footage and images from different cameras with and without fires outbreaks happening. Combining the original images given, our team had almost 7.600 images of 1920 x 1080 size (day images without fires outbreaks, day images with fires and some night images (around 16%)) to start labeling.

Samples from the datasets

 

To add even more images, Gary Diana built an algorithm to successfully extract images from the footage and at the same time avoiding the generation of images with the same landscape among them (de-duplication). This initiative added another 1.150 images of 1280 x 720 size to our dataset.

 

#3 Labeling with Labelbox

Having the datasets prepared for labeling, we gathered around 20 people dedicated to the task, created the environments on Labelbox, which is the best tool available for computer vision by allowing labeling data, managing its quality and operating a production training data pipeline, and then, at last, we started to make tests and to label the final datasets.

I managed the task but I received a huge support of Alyona Galyeva who helped the whole team not only by labeling but also by reviewing and managing everyone´s work.

In her own words:

It always starts with a mess when a group of people collaborates on a labeling project. In our case, Labelbox saved us a lot of time and effort by not allowing multiple users to label the same data. On top of that, it made our lives easier by proposing 4 roles: Labeler, Reviewer, Team Manager, and Admin. So, nobody was able to mess with data sources, data formats, and, of course, the labels made by other people.

Labelbox interface for labeling, managing and reviewing labels

 

Having both datasets labeled, next train, validation and test files were generated by the data pipeline team.

 

#4 Building the models

From the start, the team searched and studied several top-notch papers with different techniques that could be applied to solving the problem.

The challenge team created several teams in different tasks, each one focused on trying different approaches: mobilenet, semantic segmentation, Convolutional Neural Networks (CNNs) —  from simple architectures to more sophisticated ones.

Another great testimony of this step comes from Danielle Paes Barretto:

It was inspiring to see people eager to use their skills to stop wildfires and make an impact. I tried to help in all tasks; from labeling the data to building CNN models and testing them on our dataset. We also had frequent discussions which in my opinion is one of the greatest ways of learning. All in all, it was an amazing opportunity to learn and to use my knowledge for the good while meeting great people!

In addition, different techniques were successfully applied to improve results like creating patches of different sizes on original images and training over patches, data augmentation (e.g. horizontal and vertical flipping), denoising images, etc.

 

#5 Results

The final solutions were able to reach a recall between 95% and 97% while having a false positive rate between 20% and 33%, which means that these solutions were extremely successful in catching 95% to 97% of the real fires outbreaks. While the challenge partner Sintecsys is extremely happy with the results, in our second challenge, we will improve the current models by adding night time images.

 

Learning Process

As with every challenge at Omdena, it was a rich and epic journey of learning.

There is no tool more powerful to learn than getting your hands dirty in the real world.

In the words of Collaborator, Iliana Vargas:

When I heard about Omdena, I did not think twice and I applied for the challenge.The experience I had in the project in general was very gratifying for me, not only in the technical part, but also in terms of being part of a community. We had an excellent team of professionals, but above all, we had people willing to contribute with their knowledge for a project that has a social benefit.

 

What´s next?

The next natural step is to improve the model and achieve even better results. A path of building solid cutting edge technology that is not only strengthening Sintecsys position but will also allow it to move even further in their business model and value proposition.

As a community at Omdena, I am excited to build a better world, move the human spirit forward and help organizations to build AI solutions for real problems.

 

 

 

 

 

The Collaborators

I wouldn´t be able to end this article without thanking each of my colleagues in this challenge that made everything possible:

Iliana Vargas, Joon Sung Park, Sanyam Singh, Rohith Paul, Temitope Kekere, Ashish Gupta, Avikant Srivastava, Danielle Paes Barretto de Arruda Camara, Eric Massip, Kent Mok, Kritika Rupauliha, Leona Hammelrath, Nazgul Mamasheva, Nithiroj Tripatarasit, Rajashekhar Gugulothu, Rizki Fajar Nugroho, Robin Familara, Salil MishraSam Masikini, Tanya Dixit, Shaun Damon, Yash Bangera, Abhishek Unnam, Alexandr Laskorunsky, Amun Vedal, Angelo Manzatto, Billy Zhao, Carson Bentley, Lukasz Murawski, Sahand Azad, Yang Gao, Mikko Lähdeaho, Alyona Galyeva, Ana Maria Lopez Moreno, Brian Cerron, Gary Diana, Kennedy Kamande Wangari, Lasse Bøhling, Poonam Ligade, Serhiy Shekhovtsov, Kumar Mankala, François-Guillaume Fernandez and Yemissi Kifouly.

You can connect with me via LinkedIn.

 


 
 

About Omdena

Omdena is a collaborative platform where organizations work with a diverse AI community to build solutions for real world problems.

Learn more about us and Collaborative AI.

Stay in touch via our newsletter.

Be notified (a few times a month) about top-notch articles, new real-world projects, and events with our community of changemakers.

Sign up here