Abstract: An example apparatus for selecting priors includes a training set receiver to receive a training dataset. The apparatus includes a prior generator to generate a set of redundant priors based on the training dataset. The apparatus includes an intermediate trainer to train a detection CNN using the set of redundant priors. The apparatus includes a score and location receiver to send all training samples of the training dataset to the trained detection CNN and receive responses for all of the redundant priors in the set of redundant priors. The apparatus includes a subset selector to select a subset of the set of redundant priors based on the responses.

Abstract: Various embodiments are generally directed to techniques to tune a scale parameter for activations in binary neural networks, such as based on estimating a gradient for the scale parameter using quantization error, for instance. Some embodiments are particularly directed to tuning the scale parameter for activations by estimating the gradient for the scale parameter using a first “force” based on quantization error and a second, opposing, “force” based on clipping error. For instance, the first “force” based on the quantization error may give a gradient for the scale parameter that pushes the scale parameter lower to reduce the quantization error and the second “force” based on the clipping error may give a gradient for the scale parameter that moves the scale parameter higher to reduce the number of activations that are higher than a current scale parameter.

Abstract: An example apparatus for selecting priors includes a training set receiver to receive a training dataset. The apparatus includes a prior generator to generate a set of redundant priors based on the training dataset. The apparatus includes an intermediate trainer to train a detection CNN using the set of redundant priors. The apparatus includes a score and location receiver to send all training samples of the training dataset to the trained detection CNN and receive responses for all of the redundant priors in the set of redundant priors. The apparatus includes a subset selector to select a subset of the set of redundant priors based on the responses.

Abstract: A parallel physics solver may solve an equation to determine the characteristics of motion after a collision. In some embodiments, the physics solver solves the equation AX=B, where A is a sparse constrained matrix and B is the right hand vector. The sparse constrained matrix may be formed of 6×K blocks, where K is a tuning parameter that divides into the width of a single instruction multiple data processor used to implement the physics solver, without residue.

Abstract: Super-resolution images may be produced by dividing a higher resolution image into a set of non-overlapping rectangular tiles of substantially the same size. Then, each pixel in each lower resolution image is mapped to the higher resolution image and it is determined which tiles are mapped to which lower resolution image pixels. A continuous buffer may be allocated for each tile and the relevant lower resolution pixels may be stored, together with optical flow vectors, in that continuous buffer. Then, the determination of gradients may use the information now stored in the buffer to facilitate symmetric multiprocessing using multi-core processors.

Abstract: Super-resolution images may be produced by dividing a higher resolution image into a set of non-overlapping rectangular tiles of substantially the same size. Then, each pixel in each lower resolution image is mapped to the higher resolution image and it is determined which tiles are mapped to which lower resolution image pixels. A continuous buffer may be allocated for each tile and the relevant lower resolution pixels may be stored, together with optical flow vectors, in that continuous buffer. Then, the determination of gradients may use the information now stored in the buffer to facilitate symmetric multiprocessing using multi-core processors.

Abstract: Super-resolution images may be produced by dividing a higher resolution image into a set of non-overlapping rectangular tiles of substantially the same size. Then, each pixel in each lower resolution image is mapped to the higher resolution image and it is determined which tiles are mapped to which lower resolution image pixels. A continuous buffer may be allocated for each tile and the relevant lower resolution pixels may be stored, together with optical flow vectors, in that continuous buffer. Then, the determination of gradients may use the information now stored in the buffer to facilitate symmetric multiprocessing using multi-core processors.

Abstract: Estimation of a 3D layout of roads and paths traveled by pedestrians is achieved by observing the pedestrians and estimating road parameters from the pedestrian's size and position in a sequence of video frames. The system includes a foreground object detection unit to analyze video frames of a 3D scene and detect objects and object positions in video frames, an object scale prediction unit to estimate 3D transformation parameters for the objects and to predict heights of the objects based at least in part on the parameters, and a road map detection unit to estimate road boundaries of the 3D scene using the object positions to generate the road map.

Abstract: A parallel physics solver may solve an equation to determine the characteristics of motion after a collision. In some embodiments, the physics solver solves the equation AX=B, where A is a sparse constrained matrix and B is the right hand vector. The sparse constrained matrix may be formed of 6×K blocks, where K is a tuning parameter that divides into the width of a single instruction multiple data processor used to implement the physics solver, without residue.

Abstract: Super-resolution images may be produced by dividing a higher resolution image into a set of non-overlapping rectangular tiles of substantially the same size. Then, each pixel in each lower resolution image is mapped to the higher resolution image and it is determined which tiles are mapped to which lower resolution image pixels. A continuous buffer may be allocated for each tile and the relevant lower resolution pixels may be stored, together with optical flow vectors, in that continuous buffer. Then, the determination of gradients may use the information now stored in the buffer to facilitate symmetric multiprocessing using multi-core processors.

Abstract: Estimation of a 3D layout of roads and paths traveled by pedestrians is achieved by observing the pedestrians and estimating road parameters from the pedestrian's size and position in a sequence of video frames. The system includes a foreground object detection unit to analyze video frames of a 3D scene and detect objects and object positions in video frames, an object scale prediction unit to estimate 3D transformation parameters for the objects and to predict heights of the objects based at least in part on the parameters, and a road map detection unit to estimate road boundaries of the 3D scene using the object positions to generate the road map.

Abstract: Features of a mouth in a current image of a sequence of digital images of a human face may be estimated by deriving a deformable mouth model template in an iterative process. The process may include minimizing an energy function receiving iteration-dependent arguments to determine optimal transformation parameters of an iteration-dependent transformation, and transforming components of the deformable mouth model template by the iteration-dependent transformation having the optimal transformation parameters. The deformable mouth model template is initialized more than once during the process. The deformable mouth model template may include modeling of teeth.

Abstract: Features of a mouth in a current image of a sequence of digital images of a human face may be estimated by deriving a deformable mouth model template in an iterative process. The process may include minimizing an energy function receiving iteration-dependent arguments to determine optimal transformation parameters of an iteration-dependent transformation, and transforming components of the deformable mouth model template by the iteration-dependent transformation having the optimal transformation parameters. The deformable mouth model template is initialized more than once during the process. The deformable mouth model template may include modeling of teeth.

Abstract: A medium bears a deformable model configured to enable a machine to estimate positions of four points defined by X and Y coordinates, each of the points representing a facial element position in a digital image.