Two Minute Video:


Influenced by the growing plight of refugees worldwide, our team has reflected on how we can assist both refugees and the government optimize the resettlement process and increase the number of asylum-seekers who get the help they need. Currently, refugees have trouble finding a new home country that has open borders due to strict immigration laws. Because of this, it is becoming increasingly difficult for these asylum-seekers to find open-border nations that are willing to accept them and provide them essentials to lead a good life such as jobs, resources, and stability. National Governments also lack the framework to properly assist refugees along their way to resettlement as they can never coordinate the total number of people who need assistance, causing poor planning.

The major problem is inefficient migration routes. Currently, refugees don’t know which routes to take in order to get to a new nation, and therefore during migration, they may risk their lives by entering dangerous territories or taking routes that are excessively long. Non-profit organizations often provide information to refugees in which routes to take, but they also do not have proper systems to determine which routes are the most effective. Therefore, refugees currently do not have ways to find effective migration routes.

The importance of addressing this issue was brought to light to our team due to the refugee struggle as a result of the pandemic. In 2020 alone, the UNHR (refugee division of the UN) attempted to resettle 20.7 million refugees globally, but less than 1% of these refugees were ultimately resettled. Governments often struggle to communicate in time, allocate enough resources, and plan out efficient routes for refugees.

In researching more about the resettlement process, we found that after refugees reach a country, governments struggle to find jobs that match their skillset. This lack of efficiency is detrimental to both refugees and the government as a whole, as refugees are not properly utilized.

Thus, we took a unique approach from the common hackathon project. Instead of creating an application meant for general use, we developed an application for governments and nonprofits on a large scale. AI can play a large role in international cooperation between governments and nonprofits to address refugees who would otherwise receive no support. Since governments often utilize outside developers to build applications, we believe our website fills a normally unoccupied niche, and projects like this should be encouraged in the hackathon community.

Thus, we developed Trail, an application that uses AI to assist governments and non-profits in relocating and integrating refugees into safe societies.

What it does

Trail is a unique progressive web application that assists governments and nonprofits in locating and integrating refugees into new and safe nations. We specifically aim to solve a few critical issues that nonprofits and governments face. We have two pages, a refugee page which displays visualizations of safe and efficient paths for migration, and a government and nonprofit page which helps governments relocate and integrate new refugees into their society. We ask governments to upload refugee data and migration data so we can produce new paths for different refugee groups. We would work with governments to install our models into their systems so that any new data they provide could be used to build a model.

We use an ensemble of three neural networks for our system. We use historical and governmental data that has migration paths, deaths, and other refugee data to build our model. Each model is trained on a separate dataset, and we aggregate the datasets and pass them separately to each model.

The first feature is efficient route generation using geographical data. We use our neural network, which functions similarly to an autoencoder as it has to reconstruct the input, to predict these routes and then integrate the predictions on a Google Maps frontend. This allows governments to interpret which routes refugees from distressed nations should follow. This information can then be relayed to non-profit organizations that can help guide refugees out of their origin nations. Currently, governments, refugees, and organizations don’t know the most efficient and safe routes, meaning many refugees are not able to complete their journey. Using AI, we can use historical data to predict the best routes, allowing more refugees to make safe journeys.

Moreover, non-profits and governments must estimate how many refugees will migrate to their nation over a certain period of time, which currently is difficult to estimate. We use the same AI model to additionally predict the number of refugees that will migrate to a nation in a year based on current situations as well as historical data. While this model cannot predict new developments, such as new wars or disasters, based on the current circumstances, we can accurately predict how many refugees will make a migration on a specific route. This allows governments to prepare their refugee camps more effectively, which is currently not possible.

Once refugees are in their new nation, they are often stuck at refugee camps for extended periods of time because of poor integration infrastructure. Currently, governments do not have effective ways to determine which communities to send refugees to and which jobs to recommend and assign them to. We use two separate neural networks to predict which communities and jobs refugees should be placed in their new nation. Refugees can input their information such as their family size, origin nation, and other information to be matched to a certain community and job. This allows refugees to be integrated into new communities quickly, and helps the new nation improve their economy and communities.

How we built it

After numerous hours of wireframing, conceptualizing key features, and outlining tasks, we divided the challenge amongst ourselves by assigning Ishaan to developing the UI/UX, Adithya to working on data processing and frontend, Ayaan to developing our neural networks, and Viraaj to building the backend framework.

We coded the entire app in 4 languages/frameworks: HTML, CSS, Javascript, and Python(Python3 /iPython). We deployed part of our models with Flask as our backend framework and built our models with Tensorflow, Keras, and PyTorch. We used Google Cloud Platform for training our algorithm and used the Google Maps API to build the interactive map. We used Google Cloud and PythonAnywhere for our backend. We hosted our website through Netlify and Github.

We collected data for our networks using open-source datasets of historical refugee data from the UN. We used state-of-the-art network architectures from researchers. However, since we don’t have access to a cloud GPU, we were unable to entirely deploy our network ensemble.

Challenges we ran into

The primary challenge that we ran into was integrating our interactive map with our neural network. Since the networks create path coordinates, we needed a way to interactively display these. We learned that we could use a Flask backend with Python and Javascript to push the coordinates to the Google Maps API. While we were not able to fully deploy our full models, as they are too large to deploy on free and available servers, we have developed a plan to integrate them with our web server if possible.

In addition, many of these technologies were brand-new to us. We have not used advanced AI/ML on this level in the past, and learning how to develop these models was a giant hurdle we needed to overcome. In addition, we spend hours learning how to use new frameworks such as Google Cloud Platform for training our large models.

Accomplishments we are proud of

We are incredibly proud of how our team found a distinctive yet viable solution to assisting refugees in dangerous situations relocate and improve their livelihoods. We are proud that we were able to develop some of our most advanced models so far. We are extremely proud of developing a solution that hasn't been implemented in this setting. Most importantly, we were able to achieve our goals for the weekend by finishing our app, which we are very happy with.

What we learned

Our team found it incredibly fulfilling to use our AI knowledge in a way that could effectively assist governments and refugees in distressed situations. We are proud that we were able to raise awareness about and address a serious issue in refugee relocation. Seeing how we could use our software engineering skills to impact people’s livelihoods and bring attention to a serious issue in our world was the highlight of our weekend.

From a software perspective, developing geographical-processing neural networks was our main focus this weekend. We learned how to effectively use historical UN data to train advanced ML models. We grew our web development skills and polished our data skills.

What is next for Trail

We believe that our application would be best implemented on a national government level in partnership with nonprofit organizations. We specifically want to help refugees find new places to live faster, so working with the government and nonprofit organizations would be our next step.

In terms of our application, we would love to deploy all our models on the web and streamline the process of training our models and performing repeated inference with new, incoming government data. Given that our current situation prevents us from buying a web server capable of running all those processes at once, we look forward to acquiring one that can process high-level computation. Lastly, we would like to refine our algorithms to be more accurate and train with other datasets.

Share this project: