Optpresso Work Log

Let's attempt to guess espresso pull time using images of coffee grinds. Saving coffee with CNN magic.

Optpresso Work Log

Since I would like to better understand Machine Learning, I will attempt to document my process with Optpresso.

Day 1 - Too little data, too little knowledge

Day 2 - 45 shots later…

Day 3 - Poking the beast

Day 4 - Scrutinizing the model

Day 5 - No Data, No Results

Day 6 - Thanksgiving Coffee Collection

Day 7 and 8 - Waiting on data

Day 16 - And so the model gets worse…

Day 18 - It doesn’t work!

Day 20 - K, fold it

Day 30 - Finally the model gets better!

Day 31 - Check your data before celebrating

Day 34 - No matter what I try…

Day 37 - Back to basics

Day 70 - Slow and Steady

Day 80 - Test sets!

Day 86 - Inspecting data

Day 96 - Sub 70 MSE

Day 181 - Sub 70 MSE Again!

Day 246 - Looking back

Day 274 - Boring Skepticism

Coffee Sacrificed

The following is a list of coffees that have been used for data collection. Mostly lighter roasts, so this model may be way off for darker roasts.