Abstract: Disclosed herein are systems and methods that provide an end-to-end approach for performing multi-dimensional object pose estimation in the context of machine learning models. In an implementation, processing circuitry of a suitable computer inputs image data to a machine learning model that predicts a parameterized rotation vector and a parameterized translation vector for an object in the image. Next, the processing circuitry converts the parameterized rotation vector and the parameterized translation vector into a non-parameterized rotation vector and a non-parameterized translation vector respectively. Finally, the processing circuitry updates the image data based on the non-parameterized rotation vector and the non-parameterized translation vector.

Abstract: According to an aspect, a video encoder selects a block of intermediate size from a set of block sizes for intra-prediction estimation for encoding a video signal. A set of neighboring blocks with the intermediate size are tested for combining. If the set of neighboring blocks are determined to be combinable, the video encoder selects a larger block size formed by the tested neighboring blocks for encoding. On the other hand, if the set of neighboring blocks are determined to be not combinable, the video encoder selects a smaller block size from the set of tested neighboring blocks for prediction. According to another aspect of the present disclosure, the best mode for intra-prediction is determined by first intra-predicting a block with intermediate modes in a set of modes. Then the intra-predictions are performed for the neighboring modes of at least one intermediate mode.

Abstract: Systems, methods and computer readable mediums are presented for encoding a stream of input video frames, in which the input video frames are down sampled and the down sampled frames are encoded in a first encoding pass to generate a set of first pass coded frames forming a single first pass I frame and a plurality of first pass P frames formed into first pass sub-groups of pictures (SUB-GOPs). First pass encoding statistics are generated for individual first pass SUB-GOPs, and the statistics are used to encode the input video frames in a second encoding pass to generate a set of second pass coded frames.

Abstract: A hardware accelerator is configured to perform matrix multiplication and/or additional operations to optimize keypoint matching. A sum of squared error (SSE) calculation may be determined by utilizing the hardware accelerator to perform matrix multiplication to obtain a cost matrix for two sets of keypoint descriptors from two images. The hardware accelerator may determine a best cost calculation for each keypoint in each direction, which is utilized to perform keypoint matching.

Abstract: Technology is disclosed herein to execute an inference model by a processor which includes a reshape layer. In an implementation, the reshape layer of the inference model receives an output produced by a previous layer of the inference model and inserts padding into the output, then supplies the padded output as an input to a next layer of the inference model. In an implementation, the inference model includes a stitching layer at the beginning of the inference model and an un-stitch layer at the end of the model. The stitching layer of the inference model stitches together multiple input images into an image batch and supplies the image batch as an input to a subsequent layer. The un-stitch layer receives output from a penultimate layer of the inference model and unstitches the output to produce multiple output images corresponding to the multiple input images.

Abstract: An example image feature extraction system comprises an encoder neural network having a first set of layers and a decoder neural network having a second set of layers and a third set of layers. The encoder neural network receives an input image, processes the input image through the first set of layers, and computes an encoded feature map based on the input image. The decoder neural network receives the encoded feature map, processes the encoded feature map through the second set of layers to compute a keypoint score map, and processes the encoded feature map through at least a portion of the third set of layers to compute a feature description map.

Abstract: A method comprises dividing a largest coding unit (LCU) of a picture into a plurality of motion estimation regions (MERs) having size equal to or less than a predetermined size. For one or more MERs of the plurality of MERs, a number of first motion searches are performed for determining a first quad-tree based on a cost function associated with a first plurality of prediction units (PUs) of the one or more MERs. A number of second motion searches are performed for the LCU, for determining a second quad-tree, based on the cost function associated with a second plurality of PUs of the LCU. The first quad-tree or the second quad-tree is selected for performing encoding of the picture based on a comparison of a first cost of the first quad-tree with a second cost of the second quad-tree.

Abstract: The disclosure provides a sample adaptive offset (SAO) encoder. The SAO encoder includes a statistics collection (SC) block and a rate distortion optimization (RDO) block coupled to the SC block. The SC block receives a set of deblocked pixels and a set of original pixels. The SC block categorizes each deblocked pixel of the set of deblocked pixels in at least one of a plurality of band and edge categories. The SC block estimates an error in each category as difference between a deblocked pixel of the set of deblocked pixels and corresponding original pixel of the set of original pixels. The RDO block determines a set of candidate offsets associated with each category and selects a candidate offset with a minimum RD cost. The minimum RD cost is used by a SAO type block and a decision block to generate final offsets for the SAO encoder.

Abstract: Several techniques aimed at reducing computational complexity when encoding uses bi-predictively encoded frames (B-frames) are implemented in a video encoder. In an embodiment, B-frames are not used as reference frames for encoding P-frames and other B-frames. Non-use of B-frames allows a de-blocking filter used in the video encoder to be switched off when reconstructing encoded B-frames, and use of a lower complexity filter for fractional-resolution motion search for B-frames. In another embodiment, cost functions used in motion estimation for B-frames are simplified to reduce computational complexity. In one more embodiment, fractional pixel refinement in motion search for B-frames is simplified. In yet another embodiment, predictors used in motion estimation for a macro-block in a P-frame are selected from a B-frame that uses a same reference frame as the P-frame.

Abstract: A matching accelerator in the form of a hardware accelerator configured to perform matrix multiplication and/or additional operations is used to optimize keypoint matching. An SSE calculation may be determined by utilizing the matching accelerator to perform matrix multiplication to obtain a cost matrix for two sets of keypoint descriptors from two images. The hardware accelerator may determine a best cost calculation for each keypoint in each direction, which is utilized to perform keypoint matching.

Abstract: The disclosure provides a noise filter. The noise filter includes a motion estimation (ME) engine. The ME receives a current frame and a reference frame. The current frame comprising a current block and the reference frame includes a plurality of reference blocks. The ME engine generates final motion vectors. The current block comprises a plurality of current pixels. A motion compensation unit generates a motion compensated block based on the final motion vectors and the reference frame. The motion compensated block includes a plurality of motion compensated pixels. A weighted average filter multiplies each current pixel of the plurality of current pixels and a corresponding motion compensated pixel of the plurality of motion compensated pixels with a first weight and a second weight respectively. The weighted average filter generates a filtered block. A blockiness removal unit is coupled to the weighted average filter and removes artifacts in the filtered block.

Abstract: Several methods, systems, and computer program products for quantization of video content are disclosed. In an embodiment, the method includes determining by a processing module, motion information associated with a block of video data of the video content. A degree of randomness associated with the block of video data is determined by the processing module based on the motion information. A value of a quantization parameter (QP) associated with the block of video data is modulated by a quantization module based on the determined degree of randomness.

Abstract: Techniques including receiving a distorted image from a camera disposed about a vehicle, detecting, in the distorted image, corner points associated with a target object, mapping the corner points to a distortion corrected domain based on one or more camera parameters, mapping the corner points and lines between the corner points back to a distorted domain based on the camera parameters, interpolating one or more intermediate points to generate lines between the corner points in the distortion corrected domain mapping the corner points and the lines between the corner points back to a distorted domain based on the camera parameters, and adjusting a direction of travel of the vehicle based on the located target object.

Abstract: Estimation of the ground plane of a three dimensional (3D) point cloud based modifications to the random sample consensus (RANSAC) algorithm is provided. The modifications may include applying roll and pitch constraints to the selection of random planes in the 3D point cloud, using a cost function based on the number of inliers in the random plane and the number of 3D points below the random plane in the 3D point cloud, and computing a distance threshold for the 3D point cloud that is used in determining whether or not a 3D point in the 3D point cloud is an inlier of a random plane.

Abstract: The disclosure provides a sample adaptive offset (SAO) encoder. The SAO encoder includes a statistics collection (SC) block and a rate distortion optimization (RDO) block coupled to the SC block. The SC block receives a set of deblocked pixels and a set of original pixels. The SC block categorizes each deblocked pixel of the set of deblocked pixels in at least one of a plurality of band and edge categories. The SC block estimates an error in each category as difference between a deblocked pixel of the set of deblocked pixels and corresponding original pixel of the set of original pixels. The RDO block determines a set of candidate offsets associated with each category and selects a candidate offset with a minimum RD cost. The minimum RD cost is used by a SAO type block and a decision block to generate final offsets for the SAO encoder.

Abstract: Method and system to improve the performance of a video encoder. The method includes processing an initial video signal in a front-end image pre-processor to obtain a processed video signal and processor information respecting the signal, providing the processed video signal and the processor information to a video encoder, and encoding the video signal in the video encoder according to the processor information to provide an encoded video signal for storage. The system includes a video pre-processor connectable to receive an initial video signal. The video encoder in communication with the video pre-processor receives a processed video signal and a processor information. A storage medium in communication with the video encoder stores an encoded video signal.

Abstract: According to an aspect, a video encoder selects a block of intermediate size from a set of block sizes for intra-prediction estimation for encoding a video signal. A set of neighboring blocks with the intermediate size are tested for combining. If the set of neighboring blocks are determined to be combinable, the video encoder selects a larger block size formed by the tested neighboring blocks for encoding. On the other hand, if the set of neighboring blocks are determined to be not combinable, the video encoder selects a smaller block size from the set of tested neighboring blocks for prediction. According to another aspect of the present disclosure, the best mode for intra-prediction is determined by first intra-predicting a block with intermediate modes in a set of modes. Then the intra-predictions are performed for the neighboring modes of at least one intermediate mode.

Abstract: An example image feature extraction system comprises an encoder neural network having a first set of layers and a decoder neural network having a second set of layers and a third set of layers. The encoder neural network receives an input image, processes the input image through the first set of layers, and computes an encoded feature map based on the input image. The decoder neural network receives the encoded feature map, processes the encoded feature map through the second set of layers to compute a keypoint score map, and processes the encoded feature map through at least a portion of the third set of layers to compute a feature description map.

Abstract: A technique for key-point detection, including receiving, by a machine learning model, an input image, generating a set of image features for the input image, determining, by the machine learning model, based on the set of image features, a bounding box for an object detected in the input image, the bounding box described by bounding box information, identifying, by the machine learning model, based on the set of image features and a center point of the bounding box, a plurality of key-points associated with the object, filtering the plurality of key-points based on a confidence score associated with each key-point of the plurality of key-points, and outputting coordinates of the plurality of key-points, confidence scores associated with the plurality of key-points, and the bounding box information.

Abstract: This description relates to image feature extraction. In some examples, a system can include a keypoint detector and a feature list generator. The keypoint detector can be configured to upsample a keypoint score map to produce an upsampled keypoint score map. The keypoint score map can include feature scores indicative of a likelihood of at least one feature being present at keypoints in an image. The feature list generator can be configured to identify a subset of keypoints of the keypoints in the image using the feature scores of the up sampled keypoint score map, determine descriptors for the subset of keypoints based on a feature description map, and generate a keypoint descriptor map for the image based on the determined descriptors.