LiMiT (for Light Microscope Tech)

Inspiration

What it does

We analyze microscope's temperature drift and try to detect anomalies on the data and interpolate it.

How we built it

Data preprocessing:

Categorize each microscope based on their respective user ID, then use vector vectorization to normalize mapping of the data. These are done using combination of pandas, pytz, and collections.

Data Interpolation

We use piecewise cubic hermite interpolating polynomial (Pchip), which existed in sklearn. At the beginning we tried multiple ways to go with this problem (AKIMA, RBF) but it doesnt really work.

Anomaly Detection

We use copula-based outlier detection (COPOD) from pyod library

Time-series Forecasting

We use LSTM for univariate time-series forecasting using keras and try to train a model based on it

Challenges we ran into

1. Time constraint
2. A lot of trials and errors ## Accomplishments that we're proud of

What we learned

A lot

What's next for LiMiT

Get everything to work

Built With

• dash
• python