Argus

• 

  Wildfire

• 

  Flood

• 

  Setting up disaster in the cafeteria HAHA

Inspiration

Every year we see heartbreaking news about thousands of lives getting devastated with different natural calamities like flood, wildfires e.t.c.

This incurs too much loss and the geographical data of the affected areas are never observed in the right way. This is something we wanted to fix, to replace the traditional way of surveying the areas and using the data to predict the next possible disaster.

What it does

Argus stealthily flies over the affected areas without the need of a human control, it generates very critical data based on the locality density, possible population density and monocular depth data.

The data then is sent to the dashboard for visualisations and to get the nearest safest place to move to.

How we built it

1. We started with collecting data from different aerial footages to train our Deep Learning model to classify between flood, wildfires, normal scenarios.
2. We then piped the stream from drone to the model to get the classification results on the feed.
3. The stream then goes into heatmap generation which gets the data from the monocular depth estimation data as well as the colormap we generated out of our classification model.
4. All the metadata is then pushed to the dashboard.

Challenges we ran into

One of the biggest challenge was to get the right data and model to decrease the latency of model as much as possible and hence most of our time went into finding the right model.

Accomplishments that we're proud of

With the limited amount of time we had, we have been able to prototype our idea perfectly and looking at it now gives us much wider view of how much this is going to help the communities.

What we learned

While working on Argus, we got to know that there are many more ways to look into more dangerous disaster without getting human interacted by going into the dangerous scenarios themselves.

What's next for Argus

This is just the start, we will be going forward by making the model more robust and foolproof as well as making out dashboard more friendly and easier to use.

We are also looking forward to add full support of Wildfires, Volcanic Eruptions and Hurricanes.

Built With

• autonomous-drone
• css3
• html5
• npm
• opencv
• python
• pytorch
• xylem