Flare of Hope

UPDATE May 13, 2020:
Flare is refactored and renamed as TorchExpo

Lockdown huh?

I came back to Mumbai from my native place last month and within a week after that the lockdown started! Suddenly, everything just stopped for me and I got a flashback of whatever I saw in World War Z, Chernobyl, Walking Dead, etc. (I know Contagion is more closer to current situation). First 2 weeks were full on with office work, I got hit because of not planning anything, empty-minded breaks and no vision towards achieving anything. Solution was infront of me, my Bullet Journal! I saw my habit tracker and note taking graph falling down from January to March. It ticked me off really hard. By the time I realized this, it was about March end! And that’s where the magic happened!

As soon as I was pushing myself to get back on the track, I was maintaining my PyTorch C++ tutorials and was working on some tutorials on C++ which includes Jupyter Notebook for making them interactive. I saw it crossing 500 stars and then racing towards 550! I started exploring on the plans, tutorials and changes I had for this repository which I left for 2020. One night, I found myself visualizing the diagram here and immediately jumped into this new found excitment train!

These are two beautiful pictures for very different reasons. One is from my travel to my native place Mangalore. The second captures my celebration waffle and one of the last moments before lockdown where I was enjoying the work and plans for 2020 with PyTorch community.


I immediately started exploring PyTorch Android Documentation. Also I watched PyTorch Mobile @ PyTorch Developer Conference 2019 and was blown away with the demo of the experimental release. I knew that moment, I had to build something with this!

So I spent next few days breaking into Android Studio, playing with the PyTorch Java API. And that’s how Flare was born! Check it out on Google PlayStore and the source code on GitHub!

Flare is an Android application which contains collection of machine learning experiments built using PyTorch and LibTorch Java APIs.
I started with Image Classification from models available on PyTorch Hub. I am planning to support some more tasks like Image Segmentation, Text Classification, Sentiment Analysis, etc. very soon!

Flare Screen 1 Flare Screen 2 Flare Screen 3
Available on PlayStore

I am planning to work on Flare this Summer for more releases and new tasks. With that being said, I am looking for contributors who can help me with some of the new tasks. If you are interested in machine learning, evaluating models on low-resource devices like your mobile phones or you like building Android applications, Flare is waiting for you!

I hope, we will overcome this worldwide lockdown and win over the deadly virus because I believe the best damn solutions are going to come after this period as the creative minds are working on making the best of the situation and solve problems.

Remember, “The game never stops”…

I am becoming seriously allergic to 500-pound websites.