(Local Program for data in + hardware firmware)https://github.com/assasin2gamer/CalHacks (Website with Reflex) https://github.com/mishcoder/calhacks (Intel integration) https://github.com/AKUMAR0019/calhacks
Inspiration
Our project draws inspiration from the challenge of classifying emotions, a complex task. We aim to provide a reliable and cost-effective solution for training EEGs (Electroencephalograms), which can contribute to better understanding and analysis of emotions. We used a research grade EEG headset to get reliable data and do accurate sentiment analysis.
What it does
Our project leverages machine learning to train on vast datasets of EEG data and Hume's audio analysis. We use this training to develop a sentiment analysis model specifically tailored to EEGs. EEGs capture brainwaves resulting from the neuro-physiological interactions in the brain. By employing techniques like Fast Fourier Transform (FFT) and random tree (RT) modeling of time series data, we can identify EEG characteristics associated with different emotions. Our model allows us to generate visually appealing images using TogetherAI's image generation service based on the emotion classification and even further!
How we built it
We built our product using three main components: 1) Hume for sentiment data labelling 2) CockroachDB serves as our database for storing EEG data. 3) Intel Cloud Compute to compute our model 4) TogetherAI to generate images based on the emotion classification. 5) Reflex to host our website which combines the multiple data streams into one websocket.
Challenges we ran into
We ran into a few challenges: 1) We could not figure out why, but when we pinged spesifically the CockroachDB from the Intel Compute instance, the instance would freeze. 2) Our concept of using the model to be input sources for a VR game fell through when we realized the computer we brought could not handle the processing required.
Accomplishments that we're proud of
We take pride in successfully integrating multiple services, including Intel's cloud computing resources, CockroachDB for data storage, and TogetherAI for image generation, to create a cohesive solution, leverage platform advantages to create a complete and scalable tech stack.
What's next for MindScape
In the future, we plan to refine our model further and explore additional applications, such as incorporating it into a VR game or expanding our analysis capabilities for emotion classification.
We can read your mind
The details: Using reverse referencing through bloom filters, we figured out a way to introduce entropy with specific feature matches and using advanced software and hardware noise reduction we can do something most teams cannot. Through EEG data timeseries we can estimate brain activity and through that, classify emotions among other brain activities.
Log in or sign up for Devpost to join the conversation.