Dx Vision release candidate 2 [RC2]

We have been working hard on our research efforts to bring computer vision to every dental practice.

And so far our outcome has been great, so we decided to step it up a notch. After the last release candidate (RC) we had a small data set to build out first serviceable prediction model for dental X-rays. On our latest model we had more than 600 images and 4500 data-points. We want to show you our outcome on the latest update.

Our first Dx Vision RC [Release Candidate].

After we applied our new data set and tens of thousands of steps the results look like below.

Two predictions with different step count. The precision increases overtime.

What we can clearly see is that that the new algorithm trained with a richer data set can detect more object with a good precision. We will go further with thousands of images and tens of thousands of data points for our first version of the algorithm.

Another example below.


Before:

Dx Vision RC

After:

Dx Vision RC2. Step 10K — Step 40K

This is the status of Dx Vision R&D project so far. We will release this new trained model to all our users in the coming days as a service and we will open a new platform for all the dentists and dental students to play with in the near future.

Stay tuned for Dx Vision 2.0 specifications and all that is new to come.
Hint: We will include object masks in the new release :)

Until then we wish you all the best, Alen Saqe — Team Dentem.