search

microsoft / nestle-acne-assessment

This is the source code that we developed for the collaborative project between Microsoft and Nestle Skin Health, where we developed deep learning models to assess the acne severity level based on selfie images.
70 stars 40 forks source link

Regression？ #1

Open EMsnap opened 5 years ago

EMsnap commented 5 years ago

May I ask how you labeled the data while training regression model? Just labeled every image with a score between 0 - 5, right? And then graded them using the score like grade 1.8 to level 2?

hangzh-msft commented 5 years ago

We labeled each image with integer score 0 - 5. During scoring process, the score of a image is not grounded back to integer. Score 1.8 just means that it is between level 1 and 2, a little bit more toward 2.