Object recognition method and time-of-flight object recognition circuitry
The present disclosure generally pertains to an object recognition method for time-of-flight camera data, including: recognizing a real object based on a pretrained algorithm, wherein the pretrained algorithm is trained based on time-of-flight training data, wherein the time-of-flight training data are generated based on a combination of real time-of-flight data being indicative of a background, and simulated time-of-flight data generated by applying a mask on synthetic overlay image data representing a simulated object, thereby generating a masked simulated object, the mask being generated based on the synthetic overlay image data.
1 . An object recognition method, comprising:
recognizing a real object captured by an imaging processor of a first camera based on a pretrained algorithm, wherein the pretrained algorithm is trained based on training data having depth components, wherein the training data are generated based on a combination of real data having depth components derived from interpretation of data from a second camera and being indicative of a background, and simulated second data having depth components generated by applying a mask on synthetic overlay image data representing a simulated object, thereby generating a masked simulated object, the mask being generated based on the synthetic overlay image data,
wherein the training data have depth components representing both depth image data and confidence data of an image.
2 . The object recognition method of claim 1 , wherein the mask is based on at least one of a binarization of the simulated object, an erosion of the simulated object and a blurring of the simulated object.
3 . The object recognition method of claim 1 , wherein the mask is based on an application of at least one of the following to the simulated object: a random brightness change, a uniform brightness noise, and balancing the synthetic overlay image data based on the background.
4 . The object recognition method of claim 1 , wherein the pretrained algorithm is based on at least one of a generative adversarial network, a convolutional neural network, a recurrent neural network, and a convolutional neural network in combination with a neural network with a long short-term memory.
5 . The object recognition method of claim 1 , wherein the training data having depth components further include at least one of bounding box information and pixel precise masking information.
6 . The object recognition method of claim 1 , wherein the training data having depth components represent at least one of depth image data and confidence data of an image.
7 . The object recognition method of claim 1 , wherein the training data having depth components are further based on at least one of random data augmentation and hyperparameter tuning.
8 . The object recognition method of claim 1 , wherein the real data having depth components is time-of-flight data and the simulated second data is simulated time-of-flight data.
9 . The object recognition method of claim 1 , wherein the real object includes a hand.
10 . The object recognition method of claim 9 , the method further comprising: recognizing a gesture of the hand.
11 . Object recognition circuitry for recognizing an object in camera data, configured to:
recognize a real object based on a pretrained algorithm, wherein the pretrained algorithm is trained based on training data having depth components, wherein the training data are generated based on a combination of real data having depth components being indicative of a background, and simulated second data generated by applying a mask on synthetic overlay image data representing a simulated object, thereby generating a masked simulated object, the mask being generated based on the synthetic overlay image data,
wherein the training data have depth components representing both depth image data and confidence data of an image.
12 . The object recognition circuitry of claim 11 , wherein the mask is based on at least one of a binarization of the simulated object, an erosion of the simulated object and a blurring of the simulated object.
13 . The object recognition circuitry of claim 11 , wherein the mask is based on an application of at least one of the following to the simulated object: a random brightness change, a uniform brightness noise, and balancing the synthetic overlay image data based on the background.
14 . The object recognition circuitry of claim 11 , wherein the pretrained algorithm is based on at least one of a generative adversarial network, a convolutional neural network, a recurrent neural network, and a convolutional neural network in combination with a neural network with a long short-term memory.
15 . The object recognition circuitry of claim 11 , wherein the training data having depth components further include at least one of bounding box information and pixel precise masking information.
16 . The object recognition circuitry of claim 11 , wherein the training data having depth components represent at least one of depth image data and confidence data of an image.
17 . The object recognition circuitry of claim 11 , wherein the training data having depth components are further based on at least one of random data augmentation and hyperparameter tuning.
18 . The object recognition circuitry of claim 11 , wherein the pretrained algorithm is further trained based on early stopping.
19 . The object recognition circuitry of claim 11 , wherein the real object includes a hand.
20 . The object recognition circuitry of claim 19 , further configured to:
recognize a gesture of the hand.