Inspiration
Took inspiration from the common posters we saw around the neighborhood for missing dogs, etc..
What it does
It's a platform that allows the community to contribute to ongoing investigations/cases from detectives. It's primary goal was to reduce the mental load on gather data by detectives, by utilizing the community. Witnesses can upload evidence that detectives can see, and is automatically filtered based off of relevance (although every piece of data is preserved). It provides real time data insights for the witnesses' media files by utilizing websockets and multiprocessing, and can detect things like license plates, similarities between the evidence and the detective database, and can do audio analysis.
How we built it
We used the Django framework in combination with pure HTML/CSS/JS and an sqlite3 database to build the frontend. We also used tools like websockets and ajax for a better user experience. In the backend, we used tools like pytorch, tensorflow, and opencv to do data analysis. We used an LSTM model for sentamatic text analysis, and CNN model for facial emotional analysis. LSTMs are great at fitting the flow and context of sequential information, making them ideal for sentiment analysis. CNNs are superior at recognizing patterns within spatial data, which allows them them to identify emotions from the spatial arrangement of facial features in an image.
Challenges we ran into
Integrating our models with like feedback, such as websockets was very difficult. In addition to that, it's hard to make stuff look good.
Accomplishments that we're proud of
We're proud that we finished our final product, it looks nice, and it's something we would proudly show other people we know. We attempted a very ambitious project, and we were able to meet it, and that's the best feeling of all time.
What we learned
We learned how to use concepts like websockets, and ajax, as well as integrate it with a database and ML models. We also learned how to integrate websites efficiently with Django, and how to do video analysis with ML.
What's next for INVESTigator
More accurate models More analysis, models, and features for more holistic analysis. Storing more features Explore page for community Data Summarization and Coerlation
Log in or sign up for Devpost to join the conversation.