Pebblous Project

The project’s goal is to optimize the size of large image datasets for AI training. The approach involves presenting the methodology, sharing experimental findings, and supplying code for applying it to multiple datasets. The outcomes will be showcased on the client’s Tech Blog as a data story. The Data Clinic service specializes in transforming extensive data into manageable and quantifiable forms using data imaging, somewhat resembling dimensionality reduction, but linked to synthetic data generation.

The client will furnish data imaging outcomes for particular datasets. By analyzing these images, students can identify various distributional characteristics of the data. Students will then conduct lightweighting on the dataset and data imaging results provided by the client to address the stated problem. The lightweighted dataset will be presented to the client for evaluation.