Defect detection for dental appliances
A method of automated defect detection of a dental appliance comprises receiving a plurality of images of the dental appliance, and processing the plurality of images by a processing device to determine one or more defect estimations. The method comprises processing one or more sets of defect estimations to determine one or more defect classifications, wherein each set of the one or more sets comprises the one or more defect estimations associated with a subset of the plurality of images. The method comprises determining whether the dental appliance has a defect based on the one or more defect classifications associated with the one or more sets of defect estimations.
1 . A defect detection system of dental appliances, comprising:
a platform to support a dental appliance;
a camera to generate a plurality of images of the dental appliance at a plurality of position settings about the platform, wherein each image of the plurality of images depicts a distinct region of the dental appliance;
a light source to illuminate the dental appliance during generation of the plurality of images; and
a computing device to:
process the plurality of images to determine, for each image of the plurality of images, one or more defect estimations;
generate one or more sets of defect estimations, wherein each set of the one or more sets comprises the one or more defect estimations associated with a subset of the plurality of images;
generate, for each set of defect estimations of the one or more sets of defect estimations, a feature vector usable to determine whether the set of defect estimations amounts to one or more defect classifications, wherein the feature vector comprises at least a) defect coordinates for a first defect of a first image of the subset of the plurality of images and b) numbers of defects of one or more additional images of the subset of the plurality of images that are neighboring images to the first image;
process the feature vector for each of the one or more sets of defect estimations to determine the one or more defect classifications; and
determine whether the dental appliance has a defect based on the one or more defect classifications associated with the one or more sets of defect estimations.
2 . The defect detection system of claim 1 , wherein processing of the plurality of images to determine, for each image of the plurality of images, the one or more defect estimations comprises processing the plurality of images using a first machine learning model trained to identify one or more types of defects, wherein for each image of the plurality of images the first machine learning model outputs the one or more defect estimations.
3 . The defect detection system of claim 2 , wherein processing the feature vector for each of the one or more sets of defect estimations comprises processing the feature vector using at least one of the first machine learning model or a second machine learning model that outputs the one or more defect classifications.
4 . The defect detection system of claim 3 , wherein the feature vector further comprises at least one of number of defects, defect size, or defect confidence.
5 . The defect detection system of claim 1 , wherein the computing device is further to:
for each image of the plurality of images, determine a distance between a nearest point on the dental appliance and the camera that generated the image, wherein the feature vector further comprises the determined distance for each image associated with the feature vector.
6 . The defect detection system of claim 1 , wherein the feature vector associated with the subset of the plurality of images comprises:
the defect coordinates, defect size, and a first defect confidence for the first defect of the first image;
a first number of defects, of a second image of the subset of the plurality of images that is a neighboring image to the first image, and a second defect confidence for the first number of defects; and
a second number of defects, of a third image of the subset of the plurality of images that is a neighboring image to the first image, and a third defect confidence for the second number of defects.
7 . The defect detection system of claim 3 , wherein the first machine learning model comprises a convolutional neural network, and wherein the second machine learning model comprises a support vector machine.
8 . The defect detection system of claim 1 , wherein the computing device is further to:
responsive to determining that the dental appliance has a defect, output a notice that the dental appliance is a defective orthodontic appliance.
9 . The defect detection system of claim 1 , wherein each defect estimation of the one or more defect estimations comprises at least one of a defect size, a defect location, or a confidence score for the defect estimation.
10 . The defect detection system of claim 1 , wherein each set of defect estimations comprises the one or more defect estimations associated with at least three neighboring images.
11 . The defect detection system of claim 1 , wherein the one or more defect classifications comprise at least one of a break in the dental appliance, debris on the dental appliance, or an indent in the dental appliance.
12 . The defect detection system of claim 1 , wherein the plurality of images are generated from viewpoints evenly distributed about the platform.
13 . The defect detection system of claim 1 , wherein the platform is a rotatable platform, and wherein the plurality of position settings comprise a plurality of rotation settings of the rotatable platform.
14 . The defect detection system of claim 1 , wherein the platform is a transparent platform, and wherein the light source is disposed beneath the transparent platform.
15 . The defect detection system of claim 1 , wherein the plurality of images are generated from viewpoints that are at approximately 30-degree intervals of rotation about the dental appliance.
16 . The defect detection system of claim 1 , wherein the plurality of images are generated at a plurality of viewpoints that collectively are sufficient to capture a complete picture of defects on the dental appliance.
17 . The defect detection system of claim 16 , wherein the plurality of viewpoints capture interproximal regions of the dental appliance.
18 . The defect detection system of claim 16 , wherein the plurality of viewpoints capture gingival regions of the dental appliance.
19 . The defect detection system of claim 16 , wherein the plurality of viewpoints capture one or more areas of the dental appliance that are prone to damage from removal of the dental appliance from a mold or dentition.
20 . A computer readable medium comprising instructions that, when executed by a processing device, cause the processing device to perform automated defect detection of a dental appliance based on operations comprising:
receiving a plurality of images of a dental appliance, wherein each of the plurality of images was generated from a different viewpoint of the dental appliance;
processing the plurality of images to determine, for each image of the plurality of images, one or more defect estimations;
generating one or more sets of defect estimations, wherein each set of the one or more sets comprises the one or more defect estimations associated with a subset of the plurality of images;
generating, for each set of defect estimations of the one or more sets of defect estimations, a feature vector usable to determine whether the set of defect estimations amounts to one or more defect classifications, wherein the feature vector comprises at least a) defect coordinates for a first defect of a first image of the subset of the plurality of images and b) numbers of defects of one or more additional images of the subset of the plurality of images that are neighboring images to the first image;
processing the feature vector for each of the one or more sets of defect estimations to determine the one or more defect classifications; and
determining whether the dental appliance has a defect based on the one or more defect classifications associated with the one or more sets of defect estimations.
21 . A method of performing automated defect detection of a dental appliance, comprising:
generating a plurality of images of a dental appliance using a camera, wherein each of the plurality of images is generated from a different viewpoint of the dental appliance;
processing the plurality of images by a processing device to determine, for each image of the plurality of images, one or more defect estimations;
generating one or more sets of defect estimations, wherein each set of the one or more sets comprises the one or more defect estimations associated with a subset of the plurality of images;
generating, for each set of defect estimations of the one or more sets of defect estimations, a feature vector usable to determine whether the set of defect estimations amounts to one or more defect classifications, wherein the feature vector comprises at least a) defect coordinates for a first defect of a first image of the subset of the plurality of images and b) numbers of defects of one or more additional images of the subset of the plurality of images that are neighboring images to the first image;
processing the feature vector for each of the one or more sets of defect estimations to determine the one or more defect classifications; and
determining whether the dental appliance has a defect based on the one or more defect classifications associated with the one or more sets of defect estimations.