scholarly journals Detecting Crop Circles in Google Earth Images with Mask R-CNN and YOLOv3

2021 ◽  
Vol 11 (5) ◽  
pp. 2238
Author(s):  
Mohamed Lamine Mekhalfi ◽  
Carlo Nicolò ◽  
Yakoub Bazi ◽  
Mohamad Mahmoud Al Rahhal ◽  
Eslam Al Maghayreh
Keyword(s):  
Large Scale ◽  
Google Earth ◽  
Western Desert ◽  
Desert Area ◽  
Circle Detection ◽  
Detection Problem ◽  
Training Set ◽  
Detection Rates ◽  
Learning Framework ◽  
Detection And Counting

Automatic detection and counting of crop circles in the desert can be of great use for large-scale farming as it enables easy and timely management of the farming land. However, so far, the literature remains short of relevant contributions in this regard. This letter frames the crop circles detection problem within a deep learning framework. In particular, accounting for their outstanding performance in object detection, we investigate the use of Mask R-CNN (Region Based Convolutional Neural Networks) as well as YOLOv3 (You Only Look Once) models for crop circle detection in the desert. In order to quantify the performance, we build a crop circles dataset from images extracted via Google Earth over a desert area in the East Oweinat in the South-Western Desert of Egypt. The dataset totals 2511 crop circle samples. With a small training set and a relatively large test set, plausible detection rates were obtained, scoring a precision of 1 and a recall of about 0.82 for Mask R-CNN and a precision of 0.88 and a recall of 0.94 regarding YOLOv3.

scholarly journals A Weakly Supervised Deep Learning Framework for Sorghum Head Detection and Counting

2019 ◽  
Vol 2019 ◽  
pp. 1-14 ◽  
Author(s):  
Sambuddha Ghosal ◽  
Bangyou Zheng ◽  
Scott C. Chapman ◽  
Andries B. Potgieter ◽  
David R. Jordan ◽  
...  
Keyword(s):  
Deep Learning ◽  
Large Scale ◽  
Genotypic Variation ◽  
Model Performance ◽  
Phenotypic Traits ◽  
Learning Framework ◽  
Detection And Counting ◽  
Head Detection ◽  
Low Efficiency ◽  
Weakly Supervised

The yield of cereal crops such as sorghum (Sorghum bicolor L. Moench) depends on the distribution of crop-heads in varying branching arrangements. Therefore, counting the head number per unit area is critical for plant breeders to correlate with the genotypic variation in a specific breeding field. However, measuring such phenotypic traits manually is an extremely labor-intensive process and suffers from low efficiency and human errors. Moreover, the process is almost infeasible for large-scale breeding plantations or experiments. Machine learning-based approaches like deep convolutional neural network (CNN) based object detectors are promising tools for efficient object detection and counting. However, a significant limitation of such deep learning-based approaches is that they typically require a massive amount of hand-labeled images for training, which is still a tedious process. Here, we propose an active learning inspired weakly supervised deep learning framework for sorghum head detection and counting from UAV-based images. We demonstrate that it is possible to significantly reduce human labeling effort without compromising final model performance (R2 between human count and machine count is 0.88) by using a semitrained CNN model (i.e., trained with limited labeled data) to perform synthetic annotation. In addition, we also visualize key features that the network learns. This improves trustworthiness by enabling users to better understand and trust the decisions that the trained deep learning model makes.

scholarly journals Matrix SegNet: A Practical Deep Learning Framework for Landslide Mapping from Images of Different Areas with Different Spatial Resolutions

2021 ◽  
Vol 13 (16) ◽  
pp. 3158
Author(s):  
Bo Yu ◽  
Fang Chen ◽  
Chong Xu ◽  
Lei Wang ◽  
Ning Wang
Keyword(s):  
Deep Learning ◽  
Change Detection ◽  
Large Scale ◽  
Multiple Scales ◽  
Google Earth ◽  
Radiometric Correction ◽  
Learning Framework ◽  
Aspect Ratios ◽  
Starting Point ◽  
The Matrix

Practical landslide inventory maps covering large-scale areas are essential in emergency response and geohazard analysis. Recently proposed techniques in landslide detection generally focused on landslides in pure vegetation backgrounds and image radiometric correction. There are still challenges in regard to robust methods that automatically detect landslides from images with multiple platforms and without radiometric correction. It is a significant issue in practical application. In order to detect landslides from images over different large-scale areas with different spatial resolutions, this paper proposes a two-branch Matrix SegNet to semantically segment input images by change detection. The Matrix SegNet learns landslide features in multiple scales and aspect ratios. The pre- and post- event images are captured directly from Google Earth, without radiometric correction. To evaluate the proposed framework, we conducted landslide detection in four study areas with two different spatial resolutions. Moreover, two other widely used frameworks: U-Net and SegNet, were adapted to detect landslides via the same data by change detection. The experiments show that our model improves the performance largely in terms of recall, precision, F1-score, and IOU. It is a good starting point to develop a practical, deep learning landslide detection framework for large scale application, using images from different areas, with different spatial resolutions.

scholarly journals ScaleNet - Improve CNNs through Recursively Rescaling Objects

2020 ◽  
Vol 34 (07) ◽  
pp. 11426-11433
Author(s):  
Xingyi Li ◽  
Zhongang Qi ◽  
Xiaoli Fern ◽  
Fuxin Li
Keyword(s):  
Deep Learning ◽  
Large Scale ◽  
Testing Time ◽  
Learning Models ◽  
Training Set ◽  
Scale Invariant ◽  
Learning Framework ◽  
Deep Networks ◽  
Training Sets

Deep networks are often not scale-invariant hence their performance can vary wildly if recognizable objects are at an unseen scale occurring only at testing time. In this paper, we propose ScaleNet, which recursively predicts object scale in a deep learning framework. With an explicit objective to predict the scale of objects in images, ScaleNet enables pretrained deep learning models to identify objects in the scales that are not present in their training sets. By recursively calling ScaleNet, one can generalize to very large scale changes unseen in the training set. To demonstrate the robustness of our proposed framework, we conduct experiments with pretrained as well as fine-tuned classification and detection frameworks on MNIST, CIFAR-10, and MS COCO datasets and results reveal that our proposed framework significantly boosts the performances of deep networks.

scholarly journals LARGE SCALE CITY MAPPING USING SATELLITE IMAGERY / KOSMINIŲ NUOTRAUKŲ IŠ GOOGLE EARTH, TAIKOMŲ MIESTAMS KARTOGRAFUOTI STAMBIUOJU MASTELIU, REKTIFIKAVIMAS / РЕКТИФИКАЦИЯ КОСМИЧЕСКИХ СНИМКОВ ИЗ GOOGLE EARTH ДЛЯ КРУПНОМАСШТАБНОГО КАРТОГРАФИРОВАНИЯ ГОРОДОВ

2012 ◽  
Vol 37 (4) ◽  
pp. 168-171 ◽  
Author(s):  
Birutė Ruzgienė ◽  
Qian Yi Xiang ◽  
Silvija Gečytė
Keyword(s):  
Satellite Imagery ◽  
Large Scale ◽  
Modern Technology ◽  
Google Earth ◽  
Aerial Images ◽  
Control Points ◽  
Image Rectification ◽  
Image Deformation ◽  
Map Construction ◽  
Short Period

The rectification of high resolution digital aerial images or satellite imagery employed for large scale city mapping is modern technology that needs well distributed and accurately defined control points. Digital satellite imagery, obtained using widely known software Google Earth, can be applied for accurate city map construction. The method of five control points is suggested for imagery rectification introducing the algorithm offered by Prof. Ruan Wei (tong ji University, Shanghai). Image rectification software created on the basis of the above suggested algorithm can correct image deformation with required accuracy, is reliable and keeps advantages in flexibility. Experimental research on testing the applied technology has been executed using GeoEye imagery with Google Earth builder over the city of Vilnius. Orthophoto maps at the scales of 1:1000 and 1:500 are generated referring to the methodology of five control points. Reference data and rectification results are checked comparing with those received from processing digital aerial images using a digital photogrammetry approach. The image rectification process applying the investigated method takes a short period of time (about 4-5 minutes) and uses only five control points. The accuracy of the created models satisfies requirements for large scale mapping. Santrauka Didelės skiriamosios gebos skaitmeninių nuotraukų ir kosminių nuotraukų rektifikavimas miestams kartografuoti stambiuoju masteliu yra nauja technologija. Tai atliekant būtini tikslūs ir aiškiai matomi kontroliniai taškai. Skaitmeninės kosminės nuotraukos, gautos taikant plačiai žinomą programinį paketą Google Earth, gali būti naudojamos miestams kartografuoti dideliu tikslumu. Siūloma nuotraukas rektifikuoti Penkių kontrolinių taskų metodu pagal prof. Ruan Wei (Tong Ji universitetas, Šanchajus) algoritmą. Moksliniam eksperimentui pasirinkta Vilniaus GeoEye nuotrauka iš Google Earth. 1:1000 ir 1:500 mastelio ortofotografiniai žemėlapiai sudaromi Penkių kontrolinių taškų metodu. Rektifikavimo duomenys lyginami su skaitmeninių nuotraukų apdorojimo rezultatais, gautais skaitmeninės fotogrametrijos metodu. Nuotraukų rektifikavimas Penkių kontrolinių taskų metodu atitinka kartografavimo stambiuoju masteliu reikalavimus, sumažėja laiko sąnaudos. Резюме Ректификация цифровых и космических снимков высокой резолюции для крупномасштабного картографирования является новой технологией, требующей точных и четких контрольных точек. Цифровые космические снимки, полученные с использованием широкоизвестного программного пакета Google Earth, могут применяться для точного картографирования городов. Для ректификации снимков предложен метод пяти контрольных точек с применением алгоритма проф. Ruan Wei (Университет Tong Ji, Шанхай). Для научного эксперимента использован снимок города Вильнюса GeoEye из Google Earth. Ортофотографические карты в масштабе 1:1000 и 1:500 генерируются с применением метода пяти контрольных точек. Полученные результаты и данные ректификации сравниваются с результатами цифровых снимков, полученных с применением метода цифровой фотограмметрии. Ректификация снимков с применением метода пяти контрольных точек уменьшает временные расходы и удовлетворяет требования, предъявляемые к крупномасштабному картографированию.

scholarly journals Multi-task learning based Encoder-Decoder: A comprehensive detection and diagnosis system for multi-sensor data

2021 ◽  
Vol 13 (5) ◽  
pp. 168781402110131
Author(s):  
Junfeng Wu ◽  
Li Yao ◽  
Bin Liu ◽  
Zheyuan Ding ◽  
Lei Zhang
Keyword(s):  
Anomaly Detection ◽  
Event Detection ◽  
Large Scale ◽  
Multivariate Time Series ◽  
Sensor Data ◽  
Unified Framework ◽  
Diagnosis System ◽  
Learning Framework ◽  
Task Learning ◽  
Detection And Diagnosis

As more and more sensor data have been collected, automated detection, and diagnosis systems are urgently needed to lessen the increasing monitoring burden and reduce the risk of system faults. A plethora of researches have been done on anomaly detection, event detection, anomaly diagnosis respectively. However, none of current approaches can explore all these respects in one unified framework. In this work, a Multi-Task Learning based Encoder-Decoder (MTLED) which can simultaneously detect anomalies, diagnose anomalies, and detect events is proposed. In MTLED, feature matrix is introduced so that features are extracted for each time point and point-wise anomaly detection can be realized in an end-to-end way. Anomaly diagnosis and event detection share the same feature matrix with anomaly detection in the multi-task learning framework and also provide important information for system monitoring. To train such a comprehensive detection and diagnosis system, a large-scale multivariate time series dataset which contains anomalies of multiple types is generated with simulation tools. Extensive experiments on the synthetic dataset verify the effectiveness of MTLED and its multi-task learning framework, and the evaluation on a real-world dataset demonstrates that MTLED can be used in other application scenarios through transfer learning.

scholarly journals Estimating Forest Canopy Cover by Multiscale Remote Sensing in Northeast Jiangxi, China

Land ◽  
2021 ◽  
Vol 10 (4) ◽  
pp. 433
Author(s):  
Xiaolan Huang ◽  
Weicheng Wu ◽  
Tingting Shen ◽  
Lifeng Xie ◽  
Yaozu Qin ◽  
...  
Keyword(s):  
Remote Sensing ◽  
Large Scale ◽  
Forest Canopy ◽  
Canopy Cover ◽  
Google Earth ◽  
Tree Canopy ◽  
Landsat Data ◽  
High Resolution Data ◽  
Modis Ndvi ◽  
Tree Canopy Cover

This research was focused on estimation of tree canopy cover (CC) by multiscale remote sensing in south China. The key aim is to establish the relationship between CC and woody NDVI (NDVIW) or to build a CC-NDVIW model taking northeast Jiangxi as an example. Based on field CC measurements, this research used Google Earth as a complementary source to measure CC. In total, 63 sample plots of CC were created, among which 45 were applied for modeling and the remaining 18 were employed for verification. In order to ascertain the ratio R of NDVIW to the satellite observed NDVI, a 20-year time-series MODIS NDVI dataset was utilized for decomposition to obtain the NDVIW component, and then the ratio R was calculated with the equation R = (NDVIW/NDVI) *100%, respectively, for forest (CC >60%), medium woodland (CC = 25–60%) and sparse woodland (CC 1–25%). Landsat TM and OLI images that had been orthorectified by the provider USGS were atmospherically corrected using the COST model and used to derive NDVIL. R was multiplied for the NDVIL image to extract the woody NDVI (NDVIWL) from Landsat data for each of these plots. The 45 plots of CC data were linearly fitted to the NDVIWL, and a model with CC = 103.843 NDVIW + 6.157 (R2 = 0.881) was obtained. This equation was applied to predict CC at the 18 verification plots and a good agreement was found (R2 = 0.897). This validated CC-NDVIW model was further applied to the woody NDVI of forest, medium woodland and sparse woodland derived from Landsat data for regional CC estimation. An independent group of 24 measured plots was utilized for validation of the results, and an accuracy of 83.0% was obtained. Thence, the developed model has high predictivity and is suitable for large-scale estimation of CC using high-resolution data.

scholarly journals Mol-BERT: An Effective Molecular Representation with BERT for Molecular Property Prediction

2021 ◽  
Vol 2021 ◽  
pp. 1-7
Author(s):  
Juncai Li ◽  
Xiaofei Jiang
Keyword(s):  
Deep Learning ◽  
Language Processing ◽  
Large Scale ◽  
Molecular Data ◽  
Molecular Property ◽  
Property Prediction ◽  
Learning Framework ◽  
Learning Techniques ◽  
Potential Benefits ◽  
Current Sequence

Molecular property prediction is an essential task in drug discovery. Most computational approaches with deep learning techniques either focus on designing novel molecular representation or combining with some advanced models together. However, researchers pay fewer attention to the potential benefits in massive unlabeled molecular data (e.g., ZINC). This task becomes increasingly challenging owing to the limitation of the scale of labeled data. Motivated by the recent advancements of pretrained models in natural language processing, the drug molecule can be naturally viewed as language to some extent. In this paper, we investigate how to develop the pretrained model BERT to extract useful molecular substructure information for molecular property prediction. We present a novel end-to-end deep learning framework, named Mol-BERT, that combines an effective molecular representation with pretrained BERT model tailored for molecular property prediction. Specifically, a large-scale prediction BERT model is pretrained to generate the embedding of molecular substructures, by using four million unlabeled drug SMILES (i.e., ZINC 15 and ChEMBL 27). Then, the pretrained BERT model can be fine-tuned on various molecular property prediction tasks. To examine the performance of our proposed Mol-BERT, we conduct several experiments on 4 widely used molecular datasets. In comparison to the traditional and state-of-the-art baselines, the results illustrate that our proposed Mol-BERT can outperform the current sequence-based methods and achieve at least 2% improvement on ROC-AUC score on Tox21, SIDER, and ClinTox dataset.

scholarly journals Olomouc - Possibilities of Geovisualization of the Historical City

2011 ◽  
Vol 6 ◽  
pp. 267-274
Author(s):  
Stanislav Popelka ◽  
Alžběta Brychtová
Keyword(s):  
Spatial Data ◽  
Large Scale ◽  
18Th Century ◽  
3D Visualization ◽  
3D Models ◽  
Google Earth ◽  
Point Of View ◽  
World Heritage Site ◽  
Historical Maps ◽  
The City

Olomouc, nowadays a city with 100,000 inhabitants, has always been considered as one of the most prominent Czech cities. It is a social and economical centre, which history started just about the 11th century. The present appearance of the city has its roots in the 18th century, when the city was almost razed to the ground after the Thirty years’ war and a great fire in 1709. After that, the city was rebuilt to a baroque military fortress against Prussia army. At the beginning of the 20th century the majority of the fortress was demolished. Character of the town is dominated by the large number of churches, burgher’s houses and other architecturally significant buildings, like a Holy Trinity Column, a UNESCO World Heritage Site. Aim of this project was to state the most suitable methods of visualization of spatial-temporal change in historical build-up area from the tourist’s point of view, and to design and evaluate possibilities of spatial data acquisition. There are many methods of 2D and 3D visualization which are suitable for depiction of historical and contemporary situation. In the article four approaches are discussed comparison of historical and recent pictures or photos, overlaying historical maps over the orthophoto, enhanced visualization of historical map in large scale using the third dimension and photorealistic 3D models of the same area in different ages. All mentioned methods were geolocalizated using the Google Earth environment and multimedia features were added to enhance the impression of perception. Possibilities of visualization, which were outlined above, were realized on a case study of the Olomouc city. As a source of historical data were used rapport plans of the bastion fortress from the 17th century. The accuracy of historical maps was confirmed by cartometric methods with use of the MapAnalyst software. Registration of the spatial-temporal changes information has a great potential in urban planning or realization of reconstruction and particularly in the propagation of the region and increasing the knowledge of citizens about the history of Olomouc.

scholarly journals Syntax Role for Neural Semantic Role Labeling

2021 ◽  
pp. 1-48
Author(s):  
Zuchao Li ◽  
Hai Zhao ◽  
Shexia He ◽  
Jiaxun Cai
Keyword(s):  
Argument Structure ◽  
Large Scale ◽  
Language Models ◽  
Semantic Role ◽  
Semantic Role Labeling ◽  
Empirical Survey ◽  
Learning Framework ◽  
Syntactic Information ◽  
Feature Based ◽  
Predicate Argument Structure

Abstract Semantic role labeling (SRL) is dedicated to recognizing the semantic predicate-argument structure of a sentence. Previous studies in terms of traditional models have shown syntactic information can make remarkable contributions to SRL performance; however, the necessity of syntactic information was challenged by a few recent neural SRL studies that demonstrate impressive performance without syntactic backbones and suggest that syntax information becomes much less important for neural semantic role labeling, especially when paired with recent deep neural network and large-scale pre-trained language models. Despite this notion, the neural SRL field still lacks a systematic and full investigation on the relevance of syntactic information in SRL, for both dependency and both monolingual and multilingual settings. This paper intends to quantify the importance of syntactic information for neural SRL in the deep learning framework. We introduce three typical SRL frameworks (baselines), sequence-based, tree-based, and graph-based, which are accompanied by two categories of exploiting syntactic information: syntax pruningbased and syntax feature-based. Experiments are conducted on the CoNLL-2005, 2009, and 2012 benchmarks for all languages available, and results show that neural SRL models can still benefit from syntactic information under certain conditions. Furthermore, we show the quantitative significance of syntax to neural SRL models together with a thorough empirical survey using existing models.

Sign in / Sign up

Export Citation Format

Share Document