Sonia Kastner, founder and CEO of Pano, discusses using data and machine learning to detect and locate wildfires. The podcast explores the challenges of training AI models with smoke-like scenarios and the high costs of running AI experiments. It also highlights the importance of data in disaster management and supply chains.
Pano AI uses cameras mounted on mountaintop towers and an AI model trained to detect smoke to provide early alerts to fire crews, mitigating the damage caused by wildfires.
Developing an efficient AI model to detect smoke has been challenging for Pano due to the rarity of wildfires and the presence of other smoke-like elements, requiring extensive training and experimentation to improve accuracy.
Deep dives
Pano AI: Using Data to Mitigate Natural Disasters
Pano AI is a company that aims to mitigate the damage caused by natural disasters, starting with wildfires. By mounting cameras on remote mountaintop towers, Pano sends panoramic images to an AI model trained to detect smoke. This early detection system alerts fire crews, allowing them to respond quickly and prevent fires from spreading. Pano's customers include utility companies and firefighting agencies across the Western US and Australia. The company's mission extends beyond wildfires, seeking to use data to mitigate various climate-driven disasters like floods, hurricanes, and more.
Challenges in Building the AI Model
Developing an AI model to detect smoke proved to be more difficult than anticipated. While Pano uses open-source object detection models, training the model to specifically identify smoke posed challenges due to the rarity of wildfires and the existence of other smoke-like elements such as fog, dust, and barbecue smoke. Pano had to gather, label, and train the model with both smoke and non-smoke images to improve its accuracy. Building an efficient AI infrastructure and investing in ongoing experimentation has been crucial in refining and enhancing the AI model.
Pano's Approach and Capabilities
Pano provides a comprehensive solution by deploying Pano stations with ultra high-definition security cameras on mountaintop towers. The cameras continually capture high-resolution images, which are uploaded to the cloud. Pano's AI algorithm analyzes the images, adding bounding boxes where it detects smoke. Human analysts review the alerts, ensuring accuracy. Using triangulation, Pano determines the latitude and longitude of the smoke location. Automated notifications are then sent to emergency managers, providing location information and video footage. Pano offers an annual subscription-based service, handling equipment maintenance, AI, and the Pano Intelligence Center.
Expansion and Future Goals
While starting with wildfire mitigation, Pano envisions expanding into other climate-driven natural disasters like floods, hurricanes, and extreme heat. By providing more data and insights, Pano aims to support emergency managers in decision-making during response and recovery phases. The company also sees itself contributing to informing policymakers on rebuilding efforts to create resilient infrastructure against natural disasters. Pano emphasizes the importance of data in mitigating and hardening cities, transportation, and power grids, and acknowledges the increasing need for data in the face of climate change.
Sonia Kastner is the founder and CEO of Pano. Sonia’s problem is this: How do you use data and machine learning to mitigate the damage caused by climate change?
Pano mounts cameras on remote mountaintop towers, then sends images from the cameras to an AI model trained to spot wildfire smoke. The goal is to alert fire crews early, before the fire spreads.