Travel Messenger (Team 10)

Banner
Preview

Inspiration

We were a group of friends planning to go for a trip and faced difficulties in finding a good destination to go to. We also had difficulties in finding a good place to go on a budget. This was the inspiration which led us to the creation of the bot.

What it does

The messenger bot is integrated with Facebook Page. Once you open our page Travel Messenger, it pops up a messenger window, where it greets you and asks you a couple of questions to suggest you the best holiday getaway.

You come to our page Travel Messenger.
The chat pops up and greets you with a personal message and asks you to upload some pictures.
The uploaded pictures will be sent to PyTorch for analysis.
We use a pre-trained model using torchvision package.
Get classification from the PyTorch and map with the classification list locally.
From the list of places we have under the list of classification, we select the most matched first three and suggest to the user with a thumbnail and a small introduction.

How I built it

We used PyTorch and Messenger Platform, built a REST API, tunnelled our local servers to the public and published the page.

Challenges I ran into

Getting Facebook Messenger Platform to work - creating a bot and making it to respond.
Setting up a public Python API web server.

Accomplishments that I'm proud of

Using PyTorch to quickly classify images using a pre-trained model. This was something that none of the team members had attempted before, and we found that this was actually quick and easy to implement!

What I learned

That PyTorch can be quick to pick up, but will take a long time to master. We had quite a few problems getting the messenger to work, but learnt a lot from the experience.