Abstract: A multi-party privacy-preserving machine learning system is described which has a trusted execution environment comprising at least one protected memory region. An code loader at the system loads machine learning code, received from at least one of the parties, into the protected memory region. A data uploader uploads confidential data, received from at least one of the parties, to the protected memory region. The trusted execution environment executes the machine learning code using at least one data-oblivious procedure to process the confidential data and returns the result to at least one of the parties, where a data-oblivious procedure is a process where any patterns of memory accesses, patterns of disk accesses and patterns of network accesses are such that the confidential data cannot be predicted from the patterns.

Abstract: A depth detection apparatus is described which has a memory storing raw time-of-flight sensor data received from a time-of-flight sensor. The depth detection apparatus also has a trained machine learning component having been trained using training data pairs. A training data pair comprises at least one simulated raw time-of-flight sensor data value and a corresponding simulated ground truth depth value. The trained machine learning component is configured to compute in a single stage, for an item of the stored raw time-of-flight sensor data, a depth value of a surface depicted by the item, by pushing the item through the trained machine learning component.

Abstract: A depth detection apparatus is described which has a memory storing raw time-of-flight sensor data received from a time-of-flight sensor. The depth detection apparatus also has a trained machine learning component having been trained using training data pairs. A training data pair comprises at least one simulated raw time-of-flight sensor data value and a corresponding simulated ground truth depth value. The trained machine learning component is configured to compute in a single stage, for an item of the stored raw time-of-flight sensor data, a depth value of a surface depicted by the item, by pushing the item through the trained machine learning component.

Abstract: A depth detection apparatus is described which has a memory and a computation logic. The memory stores frames of raw time-of-flight sensor data received from a time-of-flight sensor, the frames having been captured by a time-of-flight camera in the presence of motion such that different ones of the frames were captured using different locations of the camera and/or with different locations of an object in a scene depicted in the frames. The computation logic has functionality to compute a plurality of depth maps from the stream of frames, whereby each frame of raw time-of-flight sensor data contributes to more than one depth map.

Abstract: A gesture detection and recognition technique is described. In one example, a sequence of data items relating to the motion of a gesturing user is received. A selected set of data items from the sequence are tested against pre-learned threshold values, to determine a probability of the sequence representing a certain gesture. If the probability is greater than a predetermined value, then the gesture is detected, and an action taken. In examples, the tests are performed by a trained decision tree classifier. In another example, the sequence of data items can be compared to pre-learned templates, and the similarity between them determined. If the similarity for a template exceeds a threshold, a likelihood value associated with a future time for a gesture associated with that template is updated. Then, when the future time is reached, the gesture is detected if the likelihood value is greater than a predefined value.

Abstract: Examples of time-of-flight (“TOF”) simulation of multipath light phenomena are described. For example, in addition to recording light intensity for a pixel during rendering, a graphics tool records the lengths (or times) and segment counts for light paths arriving at the pixel. Such multipath information can provide a characterization of the temporal light density of light that arrives at the pixel in response to one or more pulses of light. The graphics tool can use stratification and/or priority sampling to reduce variance in recorded light path samples. Realistic, physically-accurate simulation of multipath light phenomena can, in turn, help calibrate a TOF camera so that it more accurately estimates the depths of real world objects observed using the TOF camera. Various ways to improve the process of inferring imaging conditions such as depth, reflectivity, and ambient light based on images captured using a TOF camera are also described.

Abstract: Learning image processing tasks from scene reconstructions is described where the tasks may include but are not limited to: image de-noising, image in-painting, optical flow detection, interest point detection. In various embodiments training data is generated from a 2 or higher dimensional reconstruction of a scene and from empirical images of the same scene. In an example a machine learning system learns at least one parameter of a function for performing the image processing task by using the training data. In an example, the machine learning system comprises a random decision forest. In an example, the scene reconstruction is obtained by moving an image capture apparatus in an environment where the image capture apparatus has an associated dense reconstruction and camera tracking system.

Abstract: Classification algorithm optimization is described. In an example, a classification algorithm is optimized by calculating an evaluation sequence for a set of weighted feature functions that orders the feature functions in accordance with a measure of influence on the classification algorithm. Classification thresholds are determined for each step of the evaluation sequence, which indicate whether a classification decision can be made early and the classification algorithm terminated without evaluating further feature functions. In another example, a classifier applies the weighted feature functions to previously unseen data in the order of the evaluation sequence and determines a cumulative value at each step. The cumulative value is compared to the classification thresholds at each step to determine whether a classification decision can be made early without evaluating further feature functions.

Abstract: Learning image processing tasks from scene reconstructions is described where the tasks may include but are not limited to: image de-noising, image in-painting, optical flow detection, interest point detection. In various embodiments training data is generated from a 2 or higher dimensional reconstruction of a scene and from empirical images of the same scene. In an example a machine learning system learns at least one parameter of a function for performing the image processing task by using the training data. In an example, the machine learning system comprises a random decision forest. In an example, the scene reconstruction is obtained by moving an image capture apparatus in an environment where the image capture apparatus has an associated dense reconstruction and camera tracking system.

Abstract: Classification algorithm optimization is described. In an example, a classification algorithm is optimized by calculating an evaluation sequence for a set of weighted feature functions that orders the feature functions in accordance with a measure of influence on the classification algorithm. Classification thresholds are determined for each step of the evaluation sequence, which indicate whether a classification decision can be made early and the classification algorithm terminated without evaluating further feature functions. In another example, a classifier applies the weighted feature functions to previously unseen data in the order of the evaluation sequence and determines a cumulative value at each step. The cumulative value is compared to the classification thresholds at each step to determine whether a classification decision can be made early without evaluating further feature functions.

Abstract: A gesture detection and recognition technique is described. In one example, a sequence of data items relating to the motion of a gesturing user is received. A selected set of data items from the sequence are tested against pre-learned threshold values, to determine a probability of the sequence representing a certain gesture. If the probability is greater than a predetermined value, then the gesture is detected, and an action taken. In examples, the tests are performed by a trained decision tree classifier. In another example, the sequence of data items can be compared to pre-learned templates, and the similarity between them determined. If the similarity for a template exceeds a threshold, a likelihood value associated with a future time for a gesture associated with that template is updated. Then, when the future time is reached, the gesture is detected if the likelihood value is greater than a predefined value.