Object Detection Via Flexible Anchor Generation

2021 ◽  
Author(s):  
Pengxin Ding ◽  
Huan Zhou ◽  
Jinxia Shang ◽  
Xiang Zou ◽  
Minghui Wang
Keyword(s):  
Object Detection ◽  
Detection Performance ◽  
Experimental Results ◽  
Object Size ◽  
The Difference ◽  
Simple Network

This paper designs a method that can generate anchors of various shapes for the object detection framework. This method has the characteristics of novelty and flexibility. Different from the previous anchors generated by a pre-defined manner, our anchors are generated dynamically by an anchor generator. Specially, the anchor generator is not fixed but learned from the hand-designed anchors, which means that our anchor generator is able to work well in various scenes. In the inference time, the weights of anchor generator are estimated by a simple network where the input is some hand-designed anchor. In addition, in order to make the difference between the number of positive and negative samples smaller, we use an adaptive IOU threshold related to the object size to solve this problem. At the same time, we proved that our proposed method is effective and conducted a lot of experiments on the COCO dataset. Experimental results show that after replacing the anchor generation method in the previous object detectors (such as SSD, mask RCNN, and Retinanet) with our proposed method, the detection performance of the model has been greatly improved compared to before the replacement, which proves our method is effective.

scholarly journals SCALE-AWARENESS FOR MORE ACCURATE OBJECT DETECTION USING MODIFIED SINGLE SHOT DETECTORS

2021 ◽  
Vol XLIII-B2-2021 ◽  
pp. 801-808
Author(s):  
V. Tsironis ◽  
C. Stentoumis ◽  
N. Lekkas ◽  
A. Nikopoulos
Keyword(s):  
Object Detection ◽  
Sampling Method ◽  
Apparent Size ◽  
Detection Performance ◽  
Single Shot ◽  
Object Size ◽  
The Novel ◽  
Series Of Experiments ◽  
Ablation Study ◽  
The Impact

Abstract. Object detection performance is directly related to the apparent size of the object to be detected, thus most state-of-the-art algorithms dedicate different detection heads for each object size. In this work, we propose an end-to-end pipeline to adapt a single-shot object detector (SSD) to the underlying object size distribution of the target detection domain. Our contributions are the adjustments to the detector architecture and the introduction of a novel batch sampling method. To validate the effect of our method, we chose a task-specific highly specialized object detection and classification dataset of tomato fruits that apart from bounding box information, it also contains class information for three ripening stages of each tomato fruit.More specifically, the major motivation and contributions are discussed in relation to the recent bibliography. Next, an extensive analysis of our pipeline is presented, where the concept of scale alignment is thoroughly presented along with the novel sampling method. Following the results of a series of experiments, we conclude that our pipeline significantly improves over the “off-the-shelf” base single-shot detector and its detection performance is comparable to more elaborate algorithms, especially if we measure detection performance slightly disregarding box localization. Lastly, we include a stratified ablation study in the closing sections where we measure the impact of each step along our proposed SSD adaptation pipeline.

scholarly journals Object Detection Based on Fast/Faster RCNN Employing Fully Convolutional Architectures

2018 ◽  
Vol 2018 ◽  
pp. 1-7 ◽  
Author(s):  
Yun Ren ◽  
Changren Zhu ◽  
Shunping Xiao
Keyword(s):  
Object Detection ◽  
Detection Performance ◽  
Experimental Results ◽  
Feature Classification ◽  
Convolutional Networks ◽  
Detection Systems ◽  
Feature Extractor ◽  
Fully Connected

Modern object detectors always include two major parts: a feature extractor and a feature classifier as same as traditional object detectors. The deeper and wider convolutional architectures are adopted as the feature extractor at present. However, many notable object detection systems such as Fast/Faster RCNN only consider simple fully connected layers as the feature classifier. In this paper, we declare that it is beneficial for the detection performance to elaboratively design deep convolutional networks (ConvNets) of various depths for feature classification, especially using the fully convolutional architectures. In addition, this paper also demonstrates how to employ the fully convolutional architectures in the Fast/Faster RCNN. Experimental results show that a classifier based on convolutional layer is more effective for object detection than that based on fully connected layer and that the better detection performance can be achieved by employing deeper ConvNets as the feature classifier.

scholarly journals Small Object Detection in Remote Sensing Images with Residual Feature Aggregation-Based Super-Resolution and Object Detector Network

2021 ◽  
Vol 13 (9) ◽  
pp. 1854
Author(s):  
Syed Muhammad Arsalan Bashir ◽  
Yi Wang
Keyword(s):  
Remote Sensing ◽  
Object Detection ◽  
Resolution Enhancement ◽  
Super Resolution ◽  
Detection Performance ◽  
Image Resolution ◽  
Small Object ◽  
Remote Sensing Images ◽  
Feature Aggregation ◽  
Image Super Resolution

This paper deals with detecting small objects in remote sensing images from satellites or any aerial vehicle by utilizing the concept of image super-resolution for image resolution enhancement using a deep-learning-based detection method. This paper provides a rationale for image super-resolution for small objects by improving the current super-resolution (SR) framework by incorporating a cyclic generative adversarial network (GAN) and residual feature aggregation (RFA) to improve detection performance. The novelty of the method is threefold: first, a framework is proposed, independent of the final object detector used in research, i.e., YOLOv3 could be replaced with Faster R-CNN or any object detector to perform object detection; second, a residual feature aggregation network was used in the generator, which significantly improved the detection performance as the RFA network detected complex features; and third, the whole network was transformed into a cyclic GAN. The image super-resolution cyclic GAN with RFA and YOLO as the detection network is termed as SRCGAN-RFA-YOLO, which is compared with the detection accuracies of other methods. Rigorous experiments on both satellite images and aerial images (ISPRS Potsdam, VAID, and Draper Satellite Image Chronology datasets) were performed, and the results showed that the detection performance increased by using super-resolution methods for spatial resolution enhancement; for an IoU of 0.10, AP of 0.7867 was achieved for a scale factor of 16.

scholarly journals A Note on Advantages of the Fuzzy Gabor Filter in Object and Text Detection

Symmetry ◽  
2021 ◽  
Vol 13 (4) ◽  
pp. 678
Author(s):  
Vladimir Tadic ◽  
Tatjana Loncar-Turukalo ◽  
Akos Odry ◽  
Zeljen Trpovski ◽  
Attila Toth ◽  
...  
Keyword(s):  
Object Detection ◽  
Detection Method ◽  
Gabor Filter ◽  
Low Cost ◽  
Fuzzy Optimization ◽  
Detection Performance ◽  
Text Detection ◽  
License Plate ◽  
2D Gabor Filter ◽  
Fine Tune

This note presents a fuzzy optimization of Gabor filter-based object and text detection. The derivation of a 2D Gabor filter and the guidelines for the fuzzification of the filter parameters are described. The fuzzy Gabor filter proved to be a robust text an object detection method in low-quality input images as extensively evaluated in the problem of license plate localization. The extended set of examples confirmed that the fuzzy optimized Gabor filter with adequately fuzzified parameters detected the desired license plate texture components and highly improved the object detection when compared to the classic Gabor filter. The robustness of the proposed approach was further demonstrated on other images of various origin containing text and different textures, captured using low-cost or modest quality acquisition procedures. The possibility to fine tune the fuzzification procedure to better suit certain applications offers the potential to further boost detection performance.

Simulation and experimentation of a magnetorheological brake with adjustable gap

2016 ◽  
Vol 28 (12) ◽  
pp. 1614-1626 ◽  
Author(s):  
Wan-Li Song ◽  
Dong-Heng Li ◽  
Yan Tao ◽  
Na Wang ◽  
Shi-Chao Xiu
Keyword(s):  
Magnetorheological Fluid ◽  
Structure Design ◽  
Experimental Results ◽  
Wear Property ◽  
Braking Performance ◽  
Braking Torque ◽  
The Difference ◽  
The Impact ◽  
Theoretical Analyses ◽  
Magnetostatic Simulation

The aim of this work is to investigate the effect of the small magnetorheological fluid gap on the braking performance of the magnetorheological brake. In this article, theoretical analyses of the output torque are given first, and then the operating principle and design details of the magnetorheological brake whose magnetorheological fluid gap can be altered are presented and discussed. Next, the magnetic circuit of the proposed magnetorheological brake is conducted and further followed by a magnetostatic simulation of the magnetorheological brakes with different sizes of fluid gap. A prototype of the magnetorheological brake is fabricated and a series of tests are carried out to evaluate the braking performance and torque stability, as well as the verification of the simulation results. Experimental results show that the braking torque increases with the increase in the current, and the difference for the impact of the fluid gap on braking performance is huge under different currents. The rules, which the experimental results show, have an important significance on both the improvement of structure design for magnetorheological brake and the investigation of the wear property under different fluid gaps.

scholarly journals An Experimental Analysis of Model Compression Techniques for Object Detection

2020 ◽  
Author(s):  
Andrey De Aguiar Salvi ◽  
Rodrigo Coelho Barros
Keyword(s):  
Object Detection ◽  
Experimental Analysis ◽  
State Of The Art ◽  
Neural Architecture ◽  
Model Compression ◽  
Processing Power ◽  
Benchmark Datasets ◽  
The Difference ◽  
And Performance ◽  
Consumption Constraints

Recent research on Convolutional Neural Networks focuses on how to create models with a reduced number of parameters and a smaller storage size while keeping the model’s ability to perform its task, allowing the use of the best CNN for automating tasks in limited devices, with reduced processing power, memory, or energy consumption constraints. There are many different approaches in the literature: removing parameters, reduction of the floating-point precision, creating smaller models that mimic larger models, neural architecture search (NAS), etc. With all those possibilities, it is challenging to say which approach provides a better trade-off between model reduction and performance, due to the difference between the approaches, their respective models, the benchmark datasets, or variations in training details. Therefore, this article contributes to the literature by comparing three state-of-the-art model compression approaches to reduce a well-known convolutional approach for object detection, namely YOLOv3. Our experimental analysis shows that it is possible to create a reduced version of YOLOv3 with 90% fewer parameters and still outperform the original model by pruning parameters. We also create models that require only 0.43% of the original model’s inference effort.

An Effect of Air Content on the Occurrence of Cavitation

1960 ◽  
Vol 82 (4) ◽  
pp. 941-945 ◽  
Author(s):  
J. W. Holl
Keyword(s):  
Pressure Coefficient ◽  
Cavitation Number ◽  
Experimental Results ◽  
Simultaneous Occurrence ◽  
Minimum Pressure ◽  
Air Content ◽  
The Difference ◽  
Dissolved Air

The simultaneous occurrence of vaporous and gaseous cavitation on hydrofoils is considered. The experimental results show that gaseous cavitation occurs at much higher ambient pressures than that for the vaporous cavitation resulting in desinent-cavitation numbers twice the minimum-pressure coefficient of the hydrofoil. The analysis indicates that the difference between the desinent-cavitation number for the gaseous cavitation and that for the vaporous cavitation is proportional to the dissolved air content and inversely proportional to the square of the velocity.

scholarly journals Experimental Study of Vibrational Acceleration Spread and Comparison Using Three Citrus Canopy Shaker Shaking Tines

2017 ◽  
Vol 2017 ◽  
pp. 1-9 ◽  
Author(s):  
Tian-Hu Liu ◽  
Reza Ehsani ◽  
Arash Toudeshki ◽  
Xiang-Jun Zou ◽  
Hong-Jun Wang
Keyword(s):  
Experimental Study ◽  
Experimental Results ◽  
Average Decrease ◽  
Vibrational Acceleration ◽  
The Difference ◽  
Canopy Shaker

The goal of this article is to experimentally study how the vibrational acceleration spreads along the branch shaken by PVC tine, steel tine, and nylon tine for citrus canopy shaking harvesting and to compare the difference. PVC tine and steel tine have potential to be used as shaking rod for citrus canopy shaking harvesting. Nylon tine is a commonly used shaking rod. A tractor-mounted canopy shaker was developed to do the trial. The shaking frequency was set at 2.5 and 5 Hz. Experimental results showed that the vibrational acceleration at the shaking spot is not the highest. Spreading from shaking spot to the stem, it increases evidently. When spreading from stems of the outside subbranch to stems of the nearest inside subbranch, its average decrease percentage is 42%. The overall vibrational acceleration of shaking at 5 Hz is 1.85 times as high as shaking at 2.5 Hz. The overall vibrational acceleration exerted by straight PVC tine and steel tine is 1.77 and 1.97 times as high as that exerted by straight nylon tine, respectively. It is indicated that replacing nylon tine with steel tine or PVC tine helps remove the fruits inside the canopy. Replacing with steel tine is more effective than with PVC tine.

Sign in / Sign up

Export Citation Format

Share Document