Machine Learning
<-Return to all blogs
Transfer Learning from RGB to Multi-band Imagery
We used transfer learning to teach a model to take advantage of multi-band imagery without discarding the original RGB pre-training. This resulted in significant performance improvement.
Using the Dedupe Machine Learning Library for Cleaning and Matching Data
To deal with issues of apparel facility list data quality and scale quickly and efficiently we need a machine learning tool that can capture the knowledge of domain experts, find commonalities in jumbled text, and confidently compare large lists without the need to compare each individual entry.
Using Noisy Labels to Train Deep Learning Models on Satellite Imagery
How do noisy labels affect the accuracy of a deep learning model? We added different amounts of noise to the SpaceNet Vegas buildings dataset and trained some models to find out.
An Introduction to Machine Learning Accuracy Metrics
How accurate of a machine learning model can you make? It depends on how we decide to define “accurate”.