FindingHope

Landing page
About
Google authentication
Upload form
Find form
Project Pipeline

Inspiration

In our increasingly interconnected world, driven by the advent of social media and the widespread use of mobile devices, the harrowing impacts of global crises, like the Israel-Palestine and Ukraine-Russia wars, continue to become more pervasive. Amid these crises, families torn apart strive to reconnect amidst the chaos. Inspired by the multifaceted potential of AI and driven by our desire to leverage it for significant humanitarian causes, we set out to create FindingHope, an innovative solution to reconnect separated families.

What it does

FindingHope is a holistic solution that utilizes advanced AI to sift through everyday photos and videos, often revealing missing individuals who unknowingly appear in the background. The novelty of this system lies in its ability to connect seemingly unrelated videos and images to paint a broader picture. An individual who may have been reported missing under tragic circumstances in one part of the world could likewise be unknowingly captured in the background of an unrelated photo or video. By uploading 4-12 close-up facial profiling pictures, friends and family can search through FindingHope's comprehensive embedding database. Here, they can find matches that could possibly reunite people with their lost loved ones.

How we built it

The core functionality of FindingHope is powered by complex AI-driven embedding technology. It develops unique embeddings for each face in uploaded images, comparing them against a vast pool of embeddings in our Azure database. Our system uses a combination of technologies - a React UI serves as the front end communicating with a Flask API that handles communication with our Azure SQL server. Our technology does not stop at facial recognition. With our convolutional geolocation model, we can now ascertain potential locations for missing individuals with some degree of accuracy. This model essentially extracts details from pictures, evaluating every pixel to infer possible geographic coordinates. Metadata attached to the pictures is a separate entity that our technology can also rely upon but isn't often available. This provides an estimate based on the image content alone, resulting in a geolocation prediction, and location clustering.

Challenges we ran into

One of the major challenges we faced was related to Google Drive and IOS iCloud Authentication; it caused difficulties in entirely connecting our user interface to the backend Flask API due to approval issues. We spent a significant amount of time debugging this issue and contacting support. Another substantial challenge was algorithm training. Given that advanced machine learning models usually demand extensive computational resources and time, sometimes days on supercomputers or potentially weeks, we had limitations on how much we could refine the system within a 24-hour Hackathon. Achieving optimal accuracy and precision for facial recognition and geolocation was challenging given the limited time. However, we managed to create a proof of concept that has the potential, with further work, to grow into a more refined product.

Accomplishments we're proud of

We are incredibly proud of what we managed to achieve in such a short span of time. Assembling a system that could potentially have real-world positive impact is not a small feat, and doing it within the time constraints of a hackathon is something we are proud of. We greatly appreciated the opportunity presented by this challenge. Despite the tight deadline, we were successful in leveraging advanced technologies to mold a concept into a functional solution.

What we learned

From conceptualization to the final product, our journey through developing FindingHope at DubHacks has been an incredible and tiring experience. We learned to weave together diverse technologies into a coherent platform while honing our collaborative skills in a team environment.

What's next for Finding Hope

We have an ambitious vision for the future of FindingHope. We aim to augment our system's capacity to predict future potential locations of missing individuals through chronological geolocation data analysis. This enhancement could have profound implications for tracking kidnapped children or abducted war victims. Further in the future, we can create partnerships with humanitarian organizations. By integrating their resources, data, and reach, we can greatly enhance the effectiveness of FindingHope, making a substantial difference in the global humanitarian landscape. We are excited about the potential of this technology to reunite separated families and restore hope.