Towards General Purpose Object Detection: Deep Dense Grid Based Object Detection

2020 ◽  
Author(s):  
Solomon Negussie Tesema ◽  
El-Bay Bourennane
Keyword(s):  
Object Detection ◽  
General Purpose ◽  
Dense Grid ◽  
Grid Based

Development of a Partitioned Coupling Analysis System for Fluid–Structure Interactions Using an In-House ISPH Code and the Adventure System

2019 ◽  
Vol 16 (04) ◽  
pp. 1843009
Author(s):  
Masao Ogino ◽  
Takuya Iwama ◽  
Mitsuteru Asai
Keyword(s):  
Finite Element ◽  
Free Surface Flow ◽  
General Purpose ◽  
Surface Flow ◽  
Coupling Analysis ◽  
Element Analysis ◽  
Fluid Structure ◽  
Fluid Analysis ◽  
Analysis System ◽  
Grid Based

In this paper, a partitioned coupling analysis system is developed for a numerical simulation of 3-dimensional fluid–structure interaction (FSI) problems, adopting an incompressible smoothed particle hydrodynamics (SPH) method for fluid dynamics involving free surface flow and the finite element method (FEM) for structural dynamics. A coupling analysis of a particle-based method and a grid-based method has been investigated. However, most of these are developed as a function-specific application software, and therefore lack versatility. Hence, to save cost in software development and maintenance, the open source software is utilized. Especially, a general-purpose finite element analysis system, named ADVENTURE, and a general-purpose coupling analysis platform, named REVOCAP_Coupler, are employed. Moreover, techniques of an interface marker on fluid–structure boundaries and a dummy mesh for fluid analysis domain are adopted to solve the problem that the REVOCAP_Coupler performs to unify two or more grid-based method codes. To verify a developed system, the dam break problem with an elastic obstacle is demonstrated, and the result is compared with the results calculated by the other methods.

scholarly journals Grid Based Spherical CNN for Object Detection from Panoramic Images

Sensors ◽  
2019 ◽  
Vol 19 (11) ◽  
pp. 2622 ◽  
Author(s):  
Dawen Yu ◽  
Shunping Ji
Keyword(s):  
Object Detection ◽  
Data Augmentation ◽  
Dimensional Space ◽  
Three Dimensional ◽  
Dimensional Sphere ◽  
Grid Map ◽  
Bounding Box ◽  
Panoramic Images ◽  
Spherical Images ◽  
Grid Based

Recently proposed spherical convolutional neural networks (SCNNs) have shown advantages over conventional planar CNNs on classifying spherical images. However, two factors hamper their application in an objection detection task. First, a convolution in S2 (a two-dimensional sphere in three-dimensional space) or SO(3) (three-dimensional special orthogonal group) space results in the loss of an object’s location. Second, overlarge bandwidth is required to preserve a small object’s information on a sphere because the S2/SO(3) convolution must be performed on the whole sphere, instead of a local image patch. In this study, we propose a novel grid-based spherical CNN (G-SCNN) for detecting objects from spherical images. According to input bandwidth, a sphere image is transformed to a conformal grid map to be the input of the S2/SO3 convolution, and an object’s bounding box is scaled to cover an adequate area of the grid map. This solves the second problem. For the first problem, we utilize a planar region proposal network (RPN) with a data augmentation strategy that increases rotation invariance. We have also created a dataset including 600 street view panoramic images captured from a vehicle-borne panoramic camera. The dataset contains 5636 objects of interest annotated with class and bounding box and is named as WHU (Wuhan University) panoramic dataset. Results on the dataset proved our grid-based method is extremely better than the original SCNN in detecting objects from spherical images, and it outperformed several mainstream object detection networks, such as Faster R-CNN and SSD.

scholarly journals The Excellent Properties of a Dense Grid-Based HOG Feature on Face Recognition Compared to Gabor and LBP

IEEE Access ◽  
2018 ◽  
Vol 6 ◽  
pp. 29306-29319 ◽  
Author(s):  
Zheng Xiang ◽  
Hengliang Tan ◽  
Wenling Ye
Keyword(s):  
Face Recognition ◽  
Dense Grid ◽  
Grid Based

scholarly journals Dynamic Multi-LiDAR Based Multiple Object Detection and Tracking

Sensors ◽  
2019 ◽  
Vol 19 (6) ◽  
pp. 1474 ◽  
Author(s):  
Muhammad Sualeh ◽  
Gon-Woo Kim
Keyword(s):  
Object Detection ◽  
Unscented Kalman Filter ◽  
State Of The Art ◽  
Autonomous Driving ◽  
Multiple Model ◽  
Object Detection And Tracking ◽  
Detection And Tracking ◽  
Multiple Object ◽  
Multiple Object Detection ◽  
Grid Based

Environmental perception plays an essential role in autonomous driving tasks and demands robustness in cluttered dynamic environments such as complex urban scenarios. In this paper, a robust Multiple Object Detection and Tracking (MODT) algorithm for a non-stationary base is presented, using multiple 3D LiDARs for perception. The merged LiDAR data is treated with an efficient MODT framework, considering the limitations of the vehicle-embedded computing environment. The ground classification is obtained through a grid-based method while considering a non-planar ground. Furthermore, unlike prior works, 3D grid-based clustering technique is developed to detect objects under elevated structures. The centroid measurements obtained from the object detection are tracked using Interactive Multiple Model-Unscented Kalman Filter-Joint Probabilistic Data Association Filter (IMM-UKF-JPDAF). IMM captures different motion patterns, UKF handles the nonlinearities of motion models, and JPDAF associates the measurements in the presence of clutter. The proposed algorithm is implemented on two slightly dissimilar platforms, giving real-time performance on embedded computers. The performance evaluation metrics by MOT16 and ground truths provided by KITTI Datasets are used for evaluations and comparison with the state-of-the-art. The experimentation on platforms and comparisons with state-of-the-art techniques suggest that the proposed framework is a feasible solution for MODT tasks.

A Study on Improvement of Dynamic Object Detection using Dense Grid Model and Anchor Model

2018 ◽  
Vol 17 (3) ◽  
pp. 98-110
Author(s):  
Borin Yun ◽  
◽  
Sun Woo Lee ◽  
Ho Kyung Choi ◽  
Sangmin Lee ◽  
...  
Keyword(s):  
Object Detection ◽  
Dynamic Object ◽  
Grid Model ◽  
Dense Grid

scholarly journals Multimodal Fusion Object Detection System for Autonomous Vehicles

2019 ◽  
Vol 141 (7) ◽  
Author(s):  
Michael Person ◽  
Mathew Jensen ◽  
Anthony O. Smith ◽  
Hector Gutierrez
Keyword(s):  
Object Detection ◽  
Autonomous Vehicles ◽  
Point Cloud ◽  
Detection System ◽  
Three Dimensional ◽  
General Purpose ◽  
Primary Objective ◽  
Multimodal Fusion ◽  
Detection Accuracy ◽  
Cloud Data

In order for autonomous vehicles to safely navigate the road ways, accurate object detection must take place before safe path planning can occur. Currently, general purpose object detection convolutional neural network (CNN) models have the highest detection accuracies of any method. However, there is a gap in the proposed detection frameworks. Specifically, those that provide high detection accuracy necessary for deployment but do not perform inference in realtime, and those that perform inference in realtime but detection accuracy is low. We propose multimodel fusion detection system (MFDS), a sensor fusion system that combines the speed of a fast image detection CNN model along with the accuracy of light detection and range (LiDAR) point cloud data through a decision tree approach. The primary objective is to bridge the tradeoff between performance and accuracy. The motivation for MFDS is to reduce the computational complexity associated with using a CNN model to extract features from an image. To improve efficiency, MFDS extracts complimentary features from the LiDAR point cloud in order to obtain comparable detection accuracy. MFDS is novel by not only using the image detections to aid three-dimensional (3D) LiDAR detection but also using the LiDAR data to jointly bolster the image detections and provide 3D detections. MFDS achieves 3.7% higher accuracy than the base CNN detection model and is able to operate at 10 Hz. Additionally, the memory requirement for MFDS is small enough to fit on the Nvidia Tx1 when deployed on an embedded device.

Sign in / Sign up

Export Citation Format

Share Document