Interactive Deep Learning for Sorting Plant Images by Visual Phenotypes
- Huimin Han,
- Ritvik Prabhu,
- Timothy Smith,
- Kshitiz Dhakal,
- Xing Wei,
- Song Li,
- Chris North
Abstract
This paper proposes an interactive system called Andromeda that enables
users to interact with machine learning models by sorting images in a
reduced dimension plot. In our system, a dimension reduction algorithm
projects the images into a 2D space representing similarities between
the images based on visual features extracted by a deep neural network.
With Andromeda, users can alter the projection by dragging a subset of
the images into groups according to their domain expertise. The
underlying machine learning model learns the new projection by
optimizing a weighted distance function in the feature space, and the
model re-projects the images accordingly. The users can explore multiple
custom projections, and can export a model for future classification
tasks. Our approach incorporates user preferences into machine learning
model construction and allows reuse of pre-trained image processing
models to accomplish new tasks based on user inputs. Using edamame pod
images as an example, we transferred a maturity based model into a model
that can classify number of seeds per pod to demonstrate the utility of
our system.