scholarly journals A System for Weeds and Crops Identification—Reaching over 10 FPS on Raspberry Pi with the Usage of MobileNets, DenseNet and Custom Modifications

Sensors ◽  
2019 ◽  
Vol 19 (17) ◽  
pp. 3787
Author(s):  
Łukasz Chechliński ◽  
Barbara Siemiątkowska ◽  
Michał Majewski
Keyword(s):  
Low Cost ◽  
Ground Truth ◽  
Semantic Segmentation ◽  
Research Area ◽  
Raspberry Pi ◽  
University Of Technology ◽  
Knowledge Distillation ◽  
Important Research Area ◽  
Crop Area ◽  
Mobile Computers

Automated weeding is an important research area in agrorobotics. Weeds can be removed mechanically or with the precise usage of herbicides. Deep Learning techniques achieved state of the art results in many computer vision tasks, however their deployment on low-cost mobile computers is still challenging. The described system contains several novelties, compared both with its previous version and related work. It is a part of a project of the automatic weeding machine, developed by the Warsaw University of Technology and MCMS Warka Ltd. Obtained models reach satisfying accuracy (detecting 47–67% of weed area, misclasifing as weed 0.1–0.9% of crop area) at over 10 FPS on the Raspberry Pi 3B+ computer. It was tested for four different plant species at different growth stadiums and lighting conditions. The system performing semantic segmentation is based on Convolutional Neural Networks. Its custom architecture combines U-Net, MobileNets, DenseNet and ResNet concepts. Amount of needed manual ground truth labels was significantly decreased by the usage of the knowledge distillation process, learning final model which mimics an ensemble of complex models on a large database of unlabeled data. Further decrease of the inference time was obtained by two custom modifications: in the usage of separable convolutions in DenseNet block and in the number of channels in each layer. In the authors’ opinion, the described novelties can be easily transferred to other agrorobotics tasks.

scholarly journals A System for Weeds and Crops Identification – Reaching over 10 FPS on Raspberry Pi with the Usage of MobileNets, DenseNet and Custom Modifications.

2019 ◽  
Author(s):  
Łukasz Chechliński ◽  
Barbara Siemiątkowska ◽  
Michał Majewski
Keyword(s):  
Low Cost ◽  
Ground Truth ◽  
Semantic Segmentation ◽  
Research Area ◽  
Raspberry Pi ◽  
University Of Technology ◽  
Complex Models ◽  
Knowledge Distillation ◽  
Important Research Area ◽  
Mobile Computers

Automated weeding is an important research area in agrorobotics. Weeds can be removed mechanically or with the precise usage of herbicides. Deep Learning techniques achieved state of the art results in many computer vision tasks, however their deployment on low-cost mobile computers is still challenging. These paper present an advanced version of the system presented in [1]. The described system contains several novelties, compared both with its previous version and related work. It is a part of a project of the automatic weeding machine, developed by Warsaw University of Technology and MCMS Warka Ltd. The obtained model reaches satisfying accuracy at over 10~FPS on the Raspberry Pi 3B+ computer. It was tested for four different plant species at different growth stadiums and lighting conditions. The system performing semantic segmentation is based on Convolutional Neural Networks. Its custom architecture mixes U-Net, MobileNets, DenseNet and ResNet concepts. Amount of needed manual ground truth labels was significantly decreased by the usage of knowledge distillation process, learning final model to mimic an ensemble of complex models on the large database of unlabeled data. Further decrease of the inference time was obtained by two custom modifications: in the usage of separable convolutions in DenseNet block and in the number of channels in each layer. In the authors’ opinion, described novelties can be easily transferred to other agrorobotics tasks.

Trilateration approaches for seamless out-/indoor GNSS and Wi-Fi smartphone positioning

2019 ◽  
Vol 13 (1) ◽  
pp. 47-61
Author(s):  
Guenther Retscher ◽  
Jonathan Kleine ◽  
Lisa Whitemore
Keyword(s):  
Inertial Sensors ◽  
Low Cost ◽  
Ground Truth ◽  
Mobile User ◽  
Location Based Services ◽  
Raspberry Pi ◽  
Indoor Environments ◽  
Point Location ◽  
Point Determination ◽  
The Relationship

Abstract More and more sensors and receivers are found nowadays in smartphones which can enable and improve positioning for Location-based Services and other navigation applications. Apart from inertial sensors, such as accelerometers, gyroscope and magnetometer, receivers for Wireless Fidelity (Wi-Fi) and GNSS signals can be employed for positioning of a mobile user. In this study, three trilateration methods for Wi-Fi positioning are investigated whereby the influence of the derivation of the relationship between the received signal strength (RSS) and the range to an Access Points (AP) are analyzed. The first approach is a straightforward resection for point determination and the second is based on the calculation of the center of gravity in a triangle of APs while weighting the received RSS. In the third method a differential approach is employed where as in Differential GNSS (DGNSS) corrections are derived and applied to the raw RSS measurements. In this Differential Wi-Fi (DWi-Fi) method, reference stations realized by low-cost Raspberry Pi units are used to model temporal RSS variations. In the experiments in this study two different indoor environments are used, one in a laboratory and the second in the entrance of an office building. The results of the second and third approach show position deviations from the ground truth of around 2 m in dependence of the geometrical point location. Furthermore, the transition between GNSS positioning outdoors and Wi-Fi localization indoors in the entrance area of the building is studied.

scholarly journals Trilateration Approaches for Indoor Wi-Fi Positioning

2019 ◽  
Vol 94 ◽  
pp. 02002
Author(s):  
Guenther Retscher ◽  
Jonathan Kleine ◽  
Lisa Whitemore
Keyword(s):  
Inertial Sensors ◽  
Low Cost ◽  
Ground Truth ◽  
Center Of Gravity ◽  
Location Based Services ◽  
Raspberry Pi ◽  
Point Location ◽  
Differential Approach ◽  
Time Modeling ◽  
Gnss Receivers

In smartphones several sensors and receivers are embedded which enable positioning in Location-based Services and other navigation applications. They include GNSS receivers and Wireless Fidelity (Wi-Fi) cards as well as inertial sensors, such as accelerometers, gyroscope and magnetometer. In this paper, indoor Wi-Fi positioning is studied based on trilateration. Three methods are investigated which are a resection, a calculation of the center of gravity point and a differential approach. The first approach is a commonly employed resection using the ranges to the Wi-Fi Access Points (APs) as radii and intersect the circles around the APs. In the second method, the center of gravity in a triangle of APs is calculated with weighting of the received signal strength (RSS) of the Wi-Fi signals. The third approach is developed by analogy to Differential GNSS (DGNSS) and therefore termed Differential Wi-Fi (DWi-Fi). Its advantage is that a real-time modeling of the temporal RSS variations and fluctuations is possible. For that purpose, reference stations realized by low-cost Raspberry Pi units are deployed which serve at the same time as APs. The experiments conducted in a laboratory and entrance of an office building showed that position deviations from the ground truth of around 2 m are achievable with the second and third method. Thereby the positioning accuracies depend mainly on the geometrical point location in the triangle of APs and reference stations and the RSS scan duration.

scholarly journals Birds Eye View Look-Up Table Estimation with Semantic Segmentation

2021 ◽  
Vol 11 (17) ◽  
pp. 8047
Author(s):  
Dongkyu Lee ◽  
Wee Peng Tay ◽  
Seok-Cheol Kee
Keyword(s):  
Pose Estimation ◽  
Low Cost ◽  
Autonomous Vehicle ◽  
Synthetic Data ◽  
Ground Truth ◽  
Image Plane ◽  
Semantic Segmentation ◽  
Single Camera ◽  
Look Up Table ◽  
Camera Pose

In this work, a study was carried out to estimate a look-up table (LUT) that converts a camera image plane to a birds eye view (BEV) plane using a single camera. The traditional camera pose estimation fields require high costs in researching and manufacturing autonomous vehicles for the future and may require pre-configured infra. This paper proposes an autonomous vehicle driving camera calibration system that is low cost and utilizes low infra. A network that outputs an image in the form of an LUT that converts the image into a BEV by estimating the camera pose under urban road driving conditions using a single camera was studied. We propose a network that predicts human-like poses from a single image. We collected synthetic data using a simulator, made BEV and LUT as ground truth, and utilized the proposed network and ground truth to train pose estimation function. In the progress, it predicts the pose by deciphering the semantic segmentation feature and increases its performance by attaching a layer that handles the overall direction of the network. The network outputs camera angle (roll/pitch/yaw) on the 3D coordinate system so that the user can monitor learning. Since the network's output is a LUT, there is no need for additional calculation, and real-time performance is improved.

scholarly journals Knowledge Distillation for Semantic Segmentation Using Channel and Spatial Correlations and Adaptive Cross Entropy

Sensors ◽  
2020 ◽  
Vol 20 (16) ◽  
pp. 4616
Author(s):  
Sangyong Park ◽  
Yong Seok Heo
Keyword(s):  
Spatial Correlation ◽  
Loss Function ◽  
Spatial Information ◽  
Ground Truth ◽  
Semantic Segmentation ◽  
Cross Entropy ◽  
Spatial Correlations ◽  
Feature Maps ◽  
Teacher Networks ◽  
Knowledge Distillation

In this paper, we propose an efficient knowledge distillation method to train light networks using heavy networks for semantic segmentation. Most semantic segmentation networks that exhibit good accuracy are based on computationally expensive networks. These networks are not suitable for mobile applications using vision sensors, because computational resources are limited in these environments. In this view, knowledge distillation, which transfers knowledge from heavy networks acting as teachers to light networks as students, is suitable methodology. Although previous knowledge distillation approaches have been proven to improve the performance of student networks, most methods have some limitations. First, they tend to use only the spatial correlation of feature maps and ignore the relational information of their channels. Second, they can transfer false knowledge when the results of the teacher networks are not perfect. To address these two problems, we propose two loss functions: a channel and spatial correlation (CSC) loss function and an adaptive cross entropy (ACE) loss function. The former computes the full relationship of both the channel and spatial information in the feature map, and the latter adaptively exploits one-hot encodings using the ground truth labels and the probability maps predicted by the teacher network. To evaluate our method, we conduct experiments on scene parsing datasets: Cityscapes and Camvid. Our method presents significantly better performance than previous methods.

scholarly journals CBM: An IoT Enabled LiDAR Sensor for In-field Crop Height and Biomass Measurements

2021 ◽  
Author(s):  
Bikram Pratap Banerjee ◽  
German Spangenberg ◽  
Surya Kant
Keyword(s):  
High Throughput ◽  
Low Cost ◽  
Ground Truth ◽  
Plant Phenotyping ◽  
Phenotypic Characterization ◽  
Raspberry Pi ◽  
Phenotypic Traits ◽  
Small Data ◽  
Crop Species ◽  
Wheat Field

Phenotypic characterization of crop genotypes is an essential yet challenging aspect of crop management and agriculture research. Digital sensing technologies are rapidly advancing plant phenotyping and speeding-up crop breeding outcomes. However, off-the-shelf sensors might not be fully applicable and suitable for agriculture research due to diversity in crop species and specific needs during plant breeding selections. Customized sensing systems with specialized sensor hardware and software architecture provide a powerful and low-cost solution. This study designed and developed a fully integrated Raspberry Pi-based LiDAR sensor named CropBioMass (CBM), enabled by internet of things to provide a complete end-to-end pipeline. The CBM is a low-cost sensor, provides high-throughput seamless data collection in field, small data footprint, injection of data onto the remote server, and automated data processing. Phenotypic traits of crop fresh biomass, dry biomass, and plant height estimated by CBM data had high correlation with ground truth manual measurements in wheat field trial. The CBM is readily applicable for high-throughput plant phenotyping, crop monitoring, and management for precision agricultural applications.

scholarly journals Biorenewable Deep Eutectic Solvent for Selective and Scalable Conversion of Furfural into Cyclopentenone Derivatives

Molecules ◽  
2018 ◽  
Vol 23 (8) ◽  
pp. 1891 ◽  
Author(s):  
Maria Di Gioia ◽  
Monica Nardi ◽  
Paola Costanzo ◽  
Antonio De Nino ◽  
Loredana Maiuolo ◽  
...  
Keyword(s):  
Low Cost ◽  
Choline Chloride ◽  
Deep Eutectic Solvent ◽  
Reaction Medium ◽  
Research Area ◽  
Reaction Yield ◽  
Important Research ◽  
Unfavorable Effect ◽  
Important Research Area ◽  
Synthetic Routes

The development of novel synthetic routes to produce bioactive compounds starting from renewable sources has become an important research area in organic and medicinal chemistry. Here, we present a low-cost procedure for the tunable and selective conversion of biomass-produced furfural to cyclopentenone derivatives using a mixture of choline chloride and urea as a biorenewable deep eutectic solvent (DES). The proposed medium is a nontoxic, biodegradable, and could be reused up to four times without any unfavorable effect on the reaction yield. The process is tunable, clean, cheap, simple and scalable and meets most of the criteria; therefore, it can be considered as an environmental sustainable process in a natural reaction medium.

scholarly journals Designing of an Advanced Compression Bioreactor with an Implementation of a Low-Cost Controlling System Connected to a Mobile Application

Processes ◽  
2021 ◽  
Vol 9 (6) ◽  
pp. 915
Author(s):  
Gözde Dursun ◽  
Muhammad Umer ◽  
Bernd Markert ◽  
Marcus Stoffel
Keyword(s):  
Mobile Application ◽  
Low Cost ◽  
Experimental Information ◽  
Raspberry Pi ◽  
Tissue Constructs ◽  
Cost Controlling ◽  
Controlling System ◽  
And Control ◽  
Tissue Models

(1) Background: Bioreactors mimic the natural environment of cells and tissues by providing a controlled micro-environment. However, their design is often expensive and complex. Herein, we have introduced the development of a low-cost compression bioreactor which enables the application of different mechanical stimulation regimes to in vitro tissue models and provides the information of applied stress and strain in real-time. (2) Methods: The compression bioreactor is designed using a mini-computer called Raspberry Pi, which is programmed to apply compressive deformation at various strains and frequencies, as well as to measure the force applied to the tissue constructs. Besides this, we have developed a mobile application connected to the bioreactor software to monitor, command, and control experiments via mobile devices. (3) Results: Cell viability results indicate that the newly designed compression bioreactor supports cell cultivation in a sterile environment without any contamination. The developed bioreactor software plots the experimental data of dynamic mechanical loading in a long-term manner, as well as stores them for further data processing. Following in vitro uniaxial compression conditioning of 3D in vitro cartilage models, chondrocyte cell migration was altered positively compared to static cultures. (4) Conclusion: The developed compression bioreactor can support the in vitro tissue model cultivation and monitor the experimental information with a low-cost controlling system and via mobile application. The highly customizable mold inside the cultivation chamber is a significant approach to solve the limited customization capability of the traditional bioreactors. Most importantly, the compression bioreactor prevents operator- and system-dependent variability between experiments by enabling a dynamic culture in a large volume for multiple numbers of in vitro tissue constructs.

Sign in / Sign up

Export Citation Format

Share Document