Object detection training based on artificially generated images

A non-transitory computer-readable medium comprising program instructions stored on the non-transitory computer-readable medium that, when executed by at least one processor, cause a computer system to:

• receive computer-generated, non-captured image data comprising a three-dimensional representation of at least one object;
• generate a two-dimensional projection from the three-dimensional representation of at least one object;
• generate a first image as a simulated image of the at least one object based on the two-dimensional projection;
• combine the first image of the at least one object with a background image to form a synthesized image;
• generate a first edgemapped image based on the synthesized image by removing texture and color information from the synthesized image, wherein the first edgemapped image comprises edges of one or more objects from the synthesized image without texture or color information of said one or more objects, and wherein the first edgemapped image has metadata associated therewith;
• apply a first neural network to the first edgemapped image to generate a first fake image based, at least in part, on the metadata by adding noise to the first edgemapped image, wherein the first fake image is representative of a simulated X-ray image;
• store the first fake image;
• receive a second image, wherein the second image is either a real image or the first fake image;
• apply a second neural network, wherein second neural network is different from the first neural network and wherein the second neural network is configured to determine whether the second image should be classified as real or fake;
• and
• transmit first data to the first neural network based on whether the second neural network accurately classified the second image as real or fake.