Experience learning in virtual world

1. A computer-implemented method of machine-learning, comprising:

obtaining:

a test dataset of digital images of scenes belonging to a test domain, and

a domain-adaptive neural network, the domain-adaptive neural network being a machine-learned neural network taught using data consisting in scenes obtained from a preprocessing of virtual scenes belonging to a training domain where the preprocessing augments photorealism of the virtual scenes, the domain-adaptive neural network being configured for inference of spatially reconfigurable objects in a digital image of a scene of the test domain in that the domain-adaptive neural network is configured for taking as input digital image of a scene of the test domain and to detect one or more spatially reconfigurable objects in the scene by, for each of the one or more spatially reconfigurable object, inferring a bounding box encompassing the spatially reconfigurable object;

determining an intermediary domain, the intermediary domain being closer to the training domain than the test domain in terms of data distributions, the determining of the intermediary domain including, for each digital image of the test dataset:

generating a corresponding virtual scene from the digital image by applying, the digital image, a virtual scene generator which is a machine-learned neural network configured for taking as input a digital image of the test domain and inferring a corresponding virtual scene from the digital image, the virtual scene generator having been learned on a dataset of scenes each comprising one or more spatially reconfigurable objects, and

blending the digital image and the corresponding virtual scene, thereby forming a scene of the intermediary domain corresponding to the digital image; and

inferring, by applying the domain-adaptive neural network, a spatially reconfigurable object from a scene of the test domain transferred on the intermediary domain by inferring a bounding box encompassing the spatially reconfigurable object.

2. The method of claim 1 , wherein the blending of the scene of the test dataset and of the generated virtual scene is a linear blending.

3. The method of claim 1 , wherein the test dataset includes real scenes.

4. The method of claim 3 , wherein each real scene of the test dataset is a real manufacturing scene includes one or more spatially reconfigurable manufacturing tools, the domain-adaptive neural network being configured for inference of spatially reconfigurable manufacturing tools in a real manufacturing scene.

5. The method of claim 1 , wherein the training domain includes a training dataset of virtual scenes each including one or more spatially reconfigurable objects.

6. The method of claim 5 , wherein each virtual scene of the dataset of virtual scenes is a virtual manufacturing scene including one or more spatially reconfigurable manufacturing tools, the domain-adaptive neural network being configured for inference of spatially reconfigurable manufacturing tools in a real manufacturing scene.

7. The method of claim 1 , wherein the data obtained from the training domain includes scenes of another intermediary domain, the domain-adaptive neural network having been taught using the another intermediary domain, the another intermediary domain being closer to the intermediary domain than the training domain in terms of data distributions.

8. A device comprising:

a non-transitory data storage medium having recorded thereon a computer program including instructions for machine-learning that when executed by a processor causes the processor to be configured to

obtain:

a test dataset of digital images of scenes belonging to a test domain, and

a domain-adaptive neural network, the domain-adaptive neural network being a machine-learned neural network taught using data consisting in scenes obtained from a preprocessing of virtual scenes belonging to a training domain where the preprocessing augments photorealism of the virtual scenes, the domain-adaptive neural network being configured for inference of spatially reconfigurable objects in a digital image of a scene of the test domain in that the domain-adaptive neural network is configured for taking as input digital image of a scene of the test domain and to detect one or more spatially reconfigurable objects in the scene by, for each of the one or more spatially reconfigurable object, inferring a bounding box encompassing the spatially reconfigurable object;

determine an intermediary domain, the intermediary domain being closer to the training domain than the test domain in terms of data distributions, the processor being configured to determine the intermediary domain by being configured to, for each digital image of the test dataset:

generate a corresponding virtual scene from the digital image by applying, the digital image, a virtual scene generator which is a machine-learned neural network configured for taking as input a digital image of the test domain and inferring a corresponding virtual scene from the digital image, the virtual scene generator having been learned on a dataset of scenes each comprising one or more spatially reconfigurable objects, and

blend the digital image and the corresponding virtual scene, thereby forming a scene of the intermediary domain corresponding to the digital image, and

infer, by applying the domain-adaptive neural network, a spatially reconfigurable object from a scene of the test domain transferred on the intermediary domain by inferring a bounding box encompassing the spatially reconfigurable object.

9. The device of claim 8 , wherein the blending of the scene of the test dataset and of the generated virtual scene is a linear blending.

10. The device of claim 8 , further comprising the processor coupled to the non-transitory data storage medium.