Pipeline (computing)

About: Pipeline (computing) is a research topic. Over the lifetime, 26760 publications have been published within this topic receiving 204305 citations. The topic is also known as: data pipeline & computational pipeline.

Low-latency event-based visual odometry

Abstract: The agility of a robotic system is ultimately limited by the speed of its processing pipeline. The use of a Dynamic Vision Sensors (DVS), a sensor producing asynchronous events as luminance changes are perceived by its pixels, makes it possible to have a sensing pipeline of a theoretical latency of a few microseconds. However, several challenges must be overcome: a DVS does not provide the grayscale value but only changes in the luminance; and because the output is composed by a sequence of events, traditional frame-based visual odometry methods are not applicable. This paper presents the first visual odometry system based on a DVS plus a normal CMOS camera to provide the absolute brightness values. The two sources of data are automatically spatiotemporally calibrated from logs taken during normal operation. We design a visual odometry method that uses the DVS events to estimate the relative displacement since the previous CMOS frame by processing each event individually. Experiments show that the rotation can be estimated with surprising accuracy, while the translation can be estimated only very noisily, because it produces few events due to very small apparent motion.

A General Pipeline for 3D Detection of Vehicles

Abstract: Autonomous driving requires 3D perception of vehicles and other objects in the in environment. Much of the current methods support 2D vehicle detection. This paper proposes a flexible pipeline to adopt any 2D detection network and fuse it with a 3D point cloud to generate 3D information with minimum changes of the 2D detection networks. To identify the 3D box, an effective model fitting algorithm is developed based on generalised car models and score maps. A two-stage convolutional neural network (CNN) is proposed to refine the detected 3D box. This pipeline is tested on the KITTI dataset using two different 2D detection networks. The 3D detection results based on these two networks are similar, demonstrating the flexibility of the proposed pipeline. The results rank second among the 3D detection algorithms, indicating its competencies in 3D detection.

MOUSETRAP: ultra-high-speed transition-signaling asynchronous pipelines

Abstract: A new asynchronous pipeline design is introduced for high-speed applications. The pipeline uses simple transparent latches in its datapath, and small latch controllers consisting of only a single gate per pipeline stage. This simple stage structure is combined with an efficient transition-signaling protocol between stages. Initial pre-layout HSPICE simulations of a 10-stage FIFO on a 16-bit wide datapath indicate throughput of 3.51 GigaHertz in 0.25 /spl mu/ CMOS, using a conservative process. This performance is competitive even with that of wave pipelines, without the accompanying problems of complex timing and much design effort. Additionally, the new pipeline gracefully and robustly adapts to variable-speed environments. The stage implementations are extended to fork and join structures, to handle more complex system architectures.

The Ames Stereo Pipeline: Automated 3D Surface Reconstruction from Orbital Imagery

Abstract: Introduction: The Mars Orbital Laser Altimeter (MOLA) has significantly advanced the study of the Martian surface by providing geologists with a highly accurate elevation map of the entire planet [1]. However, its limited resolution (463m/pixel at the equator) and localized interpolation artifacts have rendered it insufficient for detailed studies of specific sites; e.g. geologic stratification and deposition analysis, or in the case of mission planning, landing site selection. The most common technique for obtaining higherresolution digital terrain models (DTMs) is to employ stereogrammetric techniques, however the substantial number of man-hours and resources required for this aproach has meant that relatively few of these data products have reached the scientific community. To address this problem, the Intelligent Robotics Group (IRG) at NASA Ames Research Center has developed an automated stereo processing software system, the Ames Stereo Pipeline (ASP), that is capable of generating high quality DTMs from orbital imagery using a fully automated process [2]. Approach: The image processing pipeline for the ASP can be broken down as follows. First, images are pre-processed by applying the “Sign of the Laplacian of the Gaussian” (SLOG) filter introduced by Nishihara [3]. This filter is a composition of Laplacian and Gaussian filters followed by the application of a threshold step, which results in increased robustness to lighting variation in the stereo pair. For dense stereo correlation, the ASP implements a fast area based sum of absolute difference (SOAD) correlation algorithm. The resulting disparity map encodes the offsets between matching pixels in the stereo pair. Several versions of this correlator are available including a multi-scale implementation which first computes a low resolution stereo disparity map that is then refined at successively higher levels of detail until the native resolution of the source images has been reached. Further performance is gained by adaptively partitioning the stereo images into tiles to minimize the disparity search range for any given tile. A final 3D point cloud is calculated from the disparity map by computing the closest point of intersection of two rays emanating from the cameras through the matched pixels. The ASP includes several camera models that describe the geometry of various imagers including an adaptation of the linear push-broom model [4] of line-scan imagers; a geometry that is found in many modern orbiting camera platforms. Several final data products can be generated from the 3D point cloud including 3D triangle meshes (e.g. VRML models) and ortho-rectified, map projected DTMs and camera imagery. Results: Typical processing times are on the order of minutes to tens of minutes depending on the resolution of the images. A level of quality assessment and control that is useful for many applications can be achieved with minimal human involvement. The ASP is being used in existing collaborations with Malin Space Science Systems (MSSS) and the US Geological Survey (USGS) to generate DTMs from the Narrow Angle Mars Orbital Camera (MOC-NA) (Figures 1 and 2), the MRO Context Camera (CTX), the High Resolution Stereo Camera (HRSC), and the Apollo Panoramic & Metric Cameras (Figure 3). Current and Future Activities: As work on the Ames Stereo Pipeline continues, our focus will be on integration, validation, and scalability. In particular, we have begun work in the following areas: Integration with widely adopted cartographic software: The USGS Integrated Software for Imagers and Spectrometers (ISIS) package is widely used in the planetary science community for processing raw spacecraft imagery into high level data products of scientific interest such as map projected and mosaicked imagery [5,6]. We are enabling the ASP to read ISIS image files and to utilize ISIS camera models, thereby allowing scientists to prepare data for stereo processing using a familiar tool-chain and peer reviewed camera photometric and geometric calibration.

Hardware-software partitioning and pipelined scheduling of transformative applications

Abstract: Transformative applications are computation intensive applications characterized by iterative dataflow behavior. Typical examples are image processing applications like JPEG, MPEG, etc. The performance of embedded hardware-software systems that implement transformative applications can be maximized by obtaining a pipelined design. We present a tool for hardware-software partitioning and pipelined scheduling of transformative applications. The tool uses iterative partitioning and pipelined scheduling to obtain optimal partitions that satisfy the timing and area constraints. The partitioner uses a branch and bound approach with a unique objective function that minimizes the initiation interval of the final design. We present techniques for generation of good initial solution and search-space limitation for the branch and bound algorithm. A candidate partition is evaluated by generating its pipelined schedule. The scheduler uses a novel retiming heuristic that optimizes the initiation interval, number of pipeline stages, and memory requirements of the particular design alternative. We evaluate the performance of the retiming heuristic by comparing it with an existing technique. The effectiveness of the entire tool is demonstrated by a case study of the JPEG image compression algorithm. We also evaluate the run time and design quality of the tool by experimentation with synthetic graphs.