Deep Learning for Trekking

Around two years back, Bubby and me went for a small trek from Chopta to Chandrashila, North India. Even though, the distance was about 5km(3.10 miles), we camped at a place in between named Tungnath. Our plan was to get up early in the morning and to reach Chandrashila peak top before sunrise so that we can enjoy it to the fullest.

Our camp was surrounded by snow. We got up very early and left our camp around 4am. We planned to reach the top before 5 am so that we reach there before the sunrise. We left really early that there was none before us. We used the trails which were left over the last day evening. Unfortunately the snow was pretty fluffy the previous day evening and hardened over the night. We realized this half way through the snow and there was no going back. We didn’t have shoes with spikes and hence fell down a couple of times, but we hold each other every time. By our team work, we reached the half way.

Sun started to show its rays. We can see people going to the top in far away. We realized that we can’t carry forward any further and stopped. The ice wall in front of us was so steep that we couldn’t scale it with bare shoes and no ice axe or anything that can counter balance our weight. Finally, a guy from Germany helped us with a stick using which we were able to get back in trail.

I remembered this incident when I read that drones were developed to search forest trails for lost people. These drones were developed by the researchers at the University of Zurich. With these drones, missing persons can be found and rescued quickly in forests and mountain terrains where searching them might take a long time. The researchers use AI to teach the drone to independently recognize and follow trails.

The research team solved the problem using Deep Learning, a machine learning technique used to learn feature hierarchies based on artificial neural networks. More basically, deep learning is a computer algorithm that learns to solve complex tasks from a set of “training examples,” much like a brain learns from experience. This drone was able to find directions correctly in 85% of the cases.

I am so happy that AI is presenting a dream world to us so that we can live without much difficulties. But, I want AI to grow much more, at least till when I ask my phone, “Hey Nexi, where are you?” and it has to reply “I’m under your pillow!!!”