Sanjana Shah and Aditya Shah during the testing of Smart Wild Fire Sensor (Image Source: Google)
Two-Indian origin highschool students in the US came up with a novel AI-based device that could detect wildfire at an early stage and help the authorities to tame it before its vast outbreak.
California-based high schoolers, Aditya Shah and Sanjana Shah developed Smart Wild Fire Sensor using Tensorflow, Google’s open-source software library to achieve this. The Bay Area teens built the product as part of the Google challenge that requires the participant to use Artificial Intelligence for social good.
In recent times, wildfire outbreak has been rampant in California and in other parts of the US, requiring the authorities to evacuate people to avoid a calamity of large scale. The high school seniors were inspired to build the device after their own experience of being evacuated at the nick of time.
The Smart Wild Fire Sensor, built on TensorFlow was developed by the students after feeding the software library with information such as images and weather parameters for dataflow programming. They baked AI into the product to understand and compile the environmental landscape of a vast area as it is humanly impossible to physically scan vast areas, “This is where the role of the machine and artificial intelligence come in,” Aditya exclaimed while speaking to a local daily in Pune. In order to achieve the exact precision, they partnered with Google to analyse whether images from drones and satellites could aid the device in preventing wildfire.
The device also takes into consideration other crucial inputs such as moisture content in the leaves, dryness in the air, its direction, the temperature in the area and density of the foliage to arrive at the conclusion. “During the experimental stages, the device was installed on with the camera facing down. It would then capture images of the foliage surrounding the trees in the forest,” Aditya said.
However, it wasn’t easy for the students to train the device to predict the occurrence of a wildfire due to the quantity of data which was fed into the system, speaking about some of the challenges that they faced, Sanjana said, “The main challenges were to get the right set of images and making the bunch of sensors to work with one another and co-ordinate to provide the right results. We had to collect at least 500 good images for the database to achieve 89 per cent accuracy.”
Hailing the students for their commendable work for the society, Marika Olsen, Google’s corporate communications and public affairs executive said, “The Smart Wild Fire Sensor project is a good example which highlights what can be done by AI without needing a PhD in computing. We were impressed by their ingenuity and wanted to spotlight them to show what was possible when determined people with varying backgrounds are capable of engaging of AI to solve problems in our communities.”
