How it Works
Vertical Divider

​​Trace

Trace features on a few images. Even use tracings you already have.

Train

A deep neural network learns how to recognize what you've traced.
Vertical Divider

Apply

Run your model on a new image to detect complex features.

Examples

Detecting grains while ignoring twins has challenged the community for decades.
​
Model trained on 25 images in 40 minutes on a GPU.
Applied to new image in 2 seconds.
​

Conventional methods struggle with low contrast and noise in these complex features.

Model trained on 16 images in 60 minutes on a GPU.
Applied to new image in 3 seconds.
​

Detecting these cells was impossible for traditional approaches, and even some humans.

Model trained on 600 images in 20 hours on a GPU.
Applied to new image in 2 seconds.
​

Complex, overlapping nanofiber networks have little contrast with the background.

Model trained on 36 images in 40 minutes on a GPU.
Applied to new image in 1.5 seconds.
​
© 2019