Surgical simulation for training detection and classification neural networks
A set of virtual images can be generated based on one or more real images and target rendering specifications, such that the set of virtual images correspond to (for example) different rendering specifications (or combinations thereof) than do the real images. A machine-learning model can be trained using the set of virtual images. Another real image can then be processed using the trained machine-learning model. The processing can include segmenting the other real image to detect whether and/or which objects are represented (and/or a state of the object). The object data can then be used to identify (for example) a state of a procedure.
1. A computer-implemented method comprising:
identifying a set of states represented in a procedural workflow;
for each state of the set of states:
accessing one or more base images that corresponds to the state; and
generating, for each base image of the one or more base images, first image-segmentation data that indicates a presence and/or location of each of one or more objects within the base image;
identifying a set of target rendering specifications, wherein the set of target rendering specifications include, for each image-parameter variable of one or more image-parameter variables, multiple different variable values for the image-parameter variable;
generating a set of virtual images based on the set of target rendering specifications and the one or more base images, wherein, for each of the set of states, the set of virtual images includes at least one virtual image based on the base image that corresponds to the state;
generating, for each virtual image of the set of virtual images, corresponding data that includes:
an indication of the state of the set of states with which the virtual image is associated; and
second image-segmentation data that indicates a presence and/or position of each of one or more objects within the virtual image;
training a machine-learning model using the set of virtual images and corresponding data to define a set of parameter values;
accessing a real image;
processing the real image via execution of the trained machine-learning model using the set of parameter values, wherein the processing includes identifying third image-segmentation data that indicates a presence and/or position of each of one or more objects within the real image;
generating an output based on the third image-segmentation data; and
presenting or transmitting the output.
2. The method of claim 1 , wherein generating the output includes:
identifying a state, from amongst the set of sets, with which the real image corresponds based on the third image-segmentation data;
retrieving information associated with the identified state, wherein the output includes the information.
3. The method of claim 1 , wherein generating the output includes:
identifying, based on the third image-segmentation data, a graphic or text with which to use for an augmented-reality environment;
causing the graphic or text to be superimposed on an updated visual real-time presentation of an environment, the real image having been collected at the environment.
4. The method of claim 1 , wherein the one or more objects includes a set of surgical tools.
5. The method of claim 1 , wherein the machine-learning model includes a fully convolutional network adaptation or an adversarial network model.
6. The method of claim 1 , wherein the set of target rendering specifications represents:
multiple different perspectives;
multiple different camera poses; and/or
multiple different lightings.
7. The method of claim 1 , further comprising:
accessing one or more other real images, wherein the one or more other real images and the real image correspond to frames within a video signal, and wherein processing the real image via execution of the trained machine-learning model includes processing the frames within the video signal.
8. A system comprising:
one or more data processors; and
a non-transitory computer readable storage medium containing instructions which when executed on the one or more data processors, cause the one or more data processors to perform actions including:
identifying a set of states represented in a procedural workflow;
for each state of the set of states:
accessing one or more base images that corresponds to the state; and
generating, for each base image of the one or more base images, first image-segmentation data that indicates a presence and/or location of each of one or more objects within the base image;
identifying a set of target rendering specifications, wherein the set of target rendering specifications include, for each image-parameter variable of one or more image-parameter variables, multiple different variable values for the image-parameter variable;
generating a set of virtual images based on the set of target rendering specifications and the one or more base images, wherein, for each of the set of states, the set of virtual images includes at least one virtual image based on the base image that corresponds to the state;
generating, for each virtual image of the set of virtual images, corresponding data that includes:
an indication of the state of the set of states with which the virtual image is associated; and
second image-segmentation data that indicates a presence and/or position of each of one or more objects within the virtual image;
training a machine-learning model using the set of virtual images and corresponding data to define a set of parameter values;
accessing a real image;
processing the real image via execution of the trained machine-learning model using the set of parameter values, wherein the processing includes identifying third image-segmentation data that indicates a presence and/or position of each of one or more objects within the real image;
generating an output based on the third image-segmentation data; and
presenting or transmitting the output.
9. The system of claim 8 , wherein generating the output includes:
identifying a state, from amongst the set of sets, with which the real image corresponds based on the third image-segmentation data;
retrieving information associated with the identified state, wherein the output includes the information.
10. The system of claim 8 , wherein generating the output includes:
identifying, based on the third image-segmentation data, a graphic or text with which to use for an augmented-reality environment;
causing the graphic or text to be superimposed on an updated visual real-time presentation of an environment, the real image having been collected at the environment.
11. The system of claim 8 , wherein the one or more objects includes a set of surgical tools.
12. The system of claim 8 , wherein the machine-learning model includes a fully convolutional network adaptation or an adversarial network model.
13. The system of claim 8 , wherein the set of target rendering specifications represents:
multiple different perspectives;
multiple different camera poses; and/or
multiple different lightings.
14. The system of claim 8 , wherein the actions further include:
accessing one or more other real images, wherein the one or more other real images and the real image correspond to frames within a video signal, and wherein processing the real image via execution of the trained machine-learning model includes processing the frames within the video signal.
15. A computer-program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform actions including:
identifying a set of states represented in a procedural workflow;
for each state of the set of states:
accessing one or more base images that corresponds to the state; and
generating, for each base image of the one or more base images, first image-segmentation data that indicates a presence and/or location of each of one or more objects within the base image;
identifying a set of target rendering specifications, wherein the set of target rendering specifications include, for each image-parameter variable of one or more image-parameter variables, multiple different variable values for the image-parameter variable;
generating a set of virtual images based on the set of target rendering specifications and the one or more base images, wherein, for each of the set of states, the set of virtual images includes at least one virtual image based on the base image that corresponds to the state;
generating, for each virtual image of the set of virtual images, corresponding data that includes:
an indication of the state of the set of states with which the virtual image is associated; and
second image-segmentation data that indicates a presence and/or position of each of one or more objects within the virtual image;
training a machine-learning model using the set of virtual images and corresponding data to define a set of parameter values;
accessing a real image;
processing the real image via execution of the trained machine-learning model using the set of parameter values, wherein the processing includes identifying third image-segmentation data that indicates a presence and/or position of each of one or more objects within the real image;
generating an output based on the third image-segmentation data; and
presenting or transmitting the output.
16. The computer-program product of claim 15 , wherein generating the output includes:
identifying a state, from amongst the set of sets, with which the real image corresponds based on the third image-segmentation data;
retrieving information associated with the identified state, wherein the output includes the information.
17. The computer-program product of claim 15 , wherein generating the output includes:
identifying, based on the third image-segmentation data, a graphic or text with which to use for an augmented-reality environment;
causing the graphic or text to be superimposed on an updated visual real-time presentation of an environment, the real image having been collected at the environment.
18. The computer-program product of claim 15 , wherein the one or more objects includes a set of surgical tools.
19. The computer-program product of claim 15 , wherein the machine-learning model includes a fully convolutional network adaptation or an adversarial network model.
20. The computer-program product of claim 15 , wherein the set of target rendering specifications represents:
multiple different perspectives;
multiple different camera poses; and/or
multiple different lightings.