Masked Regions Research Articles

K-mer counting and curated libraries drive efficient annotation of repeats in plant genomes.

The annotation of repetitive sequences within plant genomes can help in the interpretation of observed phenotypes. Moreover, repeat masking is required for tasks such as whole-genome alignment, promoter analysis, or pangenome exploration. Although homology-based annotation methods are computationally expensive, k-mer strategies for masking are orders of magnitude faster. Here, we benchmarked a two-step approach, where repeats were first called by k-mer counting and then annotated by comparison to curated libraries. This hybrid protocol was tested on 20 plant genomes from Ensembl, with the k-mer-based Repeat Detector (Red) and two repeat libraries (REdat, last updated in 2013, and nrTEplants, curated for this work). Custom libraries produced by RepeatModeler were also tested. We obtained repeated genome fractions that matched those reported in the literature but with shorter repeated elements than those produced directly by sequence homology. Inspection of the masked regions that overlapped genes revealed no preference for specific protein domains. Most Red-masked sequences could be successfully classified by sequence similarity, with the complete protocol taking less than 2 h on a desktop Linux box. A guide to curating your own repeat libraries and the scripts for masking and annotating plant genomes can be obtained at https://github.com/Ensembl/plant-scripts.

A soft computing automatic based in deep learning with use of fine-tuning for pulmonary segmentation in computed tomography images

The segmentation of medical images is a significant challenge for computer vision and medical techniques. In the lungs, the difficulty is mainly due to the presence of lung diseases, different dimensions of the lung, and different types and configurations of medical imaging devices. Computed tomography (CT) is a tool to aid clinical diagnosis; several systems based on computer-aided diagnostics (CAD) are developed or enhanced from CT images combined with computational methods. This study proposes an innovative approach based on the generalization of models for pulmonary segmentation in CT images, using the Convolutional Neural Network Based on Mask Regions (Mask R-CNN) combined with image processing techniques, K-means clustering, region growing and Parzen window through fine-tuning. Our models achieved satisfactory results with 97% Accuracy and 98% Dice, 99% Sensitivity, and 97% Positive Predictive Value (PPV) in our best MPK model. The method was able to generalize its learning to solve different pulmonary segmentation problems in different lung CT databases. A second set of data was used, obtaining even better results. The results obtained for this second data set were: 98% Accuracy, 99% Dice, 100% Sensitivity, and 98% PPV, demonstrating the effectiveness of our method.

Recognition of Masked Face Using Convolutional Neural Network Algorithm

The covid -19 is an unparalleled crisis resulting in huge number of casualties security problem. So has to scale back the spread of corona virus, people often wear a mask to guard themselves. Indeed, during this challenging context, the matter of face recognition is usually like periocular recognition involving iris, pupil, sclera, upper and lower eyelids, eye folds, eye corners, skin texture, fine wrinkles, complexion, skin color, skin pores etc. In this paper, we propose a reliable method supported discard masked region and deep learning based features so as to deal with the matter of masked face recognition process. The primary step to discard the masked face region. Next, we apply deep learning algorithm to extract the simplest features from obtained regions (mostly eyes and forehead regions). This leads to good accuracy than the previous work for detecting the masked face.

Smart training: Mask R-CNN oriented approach

This paper is aimed at the usage of an augmented reality assisted system set up on the smart-glasses for training activities. Literature review leads us to a comparison among related technologies, yielding that Mask Regions with Convolutional Neural Network (R-CNN) oriented approach fits the study needs. The proposed method including (1) pointing gesture capture, (2) finger-pointing analysis, and (3) virtual tool positioning and rotation angle are developed. Results show that the recognition of object detection is 95.5%, the Kappa value of recognition of gesture detection is 0.93, and the average time for detecting pointing gesture is 0.26 seconds. Furthermore, even under different lighting, such as indoor and outdoor, the pointing analysis accuracy is up to 79%. The error between the analysis angle and the actual angle is only 1.32 degrees. The results proved that the system is well suited to present the effect of augmented reality, making it applicable for real world usage.

Improve of Mask R-CNN in Edge Segmentation

Nowadays, grasping robot plays an important role in many automatic systems in the industrial environment. An excellent grasping robot can detect, localize, and pick objects accurately but to perfectly achieve these tasks, it is still a challenge in the computer vision field. Especially, segmentation task, which is understood as both detection and localization, is the hardest problem. To deal with this problem, the state-of-the-art Mask Region Convolution Neural Network (Mask R-CNN) was introduced and obtained an exceptional result. But this superb model does not certainly perform well when working with harsh locations of objects. The edge and border regions are usually misunderstood as the background, this leads to the failure in localizing objects to submit a good grasping plan. Thus, in this paper, we introduce a novel method that combines the original Mask R-CNN pipeline and 3D algorithms branch to preserve and classify the edge region. This results from the improvement of the performance of Mask R-CNN in detailed segmentation. Concretely, the significant improvement practiced in harsh situations of object location was obviously discussed in the experimental result section. Both IoU and mAP indicators are increased. Specifically, mAP, which directly reflects the semantic segmentation ability of a model, raised from 0.39 to 0.46. This approach opens a better way to determine the object location and grasping plan.

A Mask R-CNN based particle identification for quantitative shape evaluation of granular materials

Particle identification and shape evaluation of granular materials from their realistic packing images are challenging and of great interest to many engineers and researchers. In this study, a systematic tool is developed based on computing techniques, including deep learning and computational geometry. First, image datasets of the target granular particles with well-labeled masks are established. The Mask Region Convolutional Neural Network (Mask R-CNN) is employed to implement the end-to-end instance segmentation and contour extraction of particles on different realistic images. Since Mask R-CNN models have several different feature extraction backbones, the optimal model is selected and then trained on the established datasets using transfer learning technique. After the particles are successfully identified from images of cobble and ballast, the elongation, angularity, and roughness are evaluated and the statistical shape analysis is conducted. The proposed method has strong generalization ability, especially for densely-packed particles.

Egocentric-View Fingertip Detection for Air Writing Based on Convolutional Neural Networks.

This research investigated real-time fingertip detection in frames captured from the increasingly popular wearable device, smart glasses. The egocentric-view fingertip detection and character recognition can be used to create a novel way of inputting texts. We first employed Unity3D to build a synthetic dataset with pointing gestures from the first-person perspective. The obvious benefits of using synthetic data are that they eliminate the need for time-consuming and error-prone manual labeling and they provide a large and high-quality dataset for a wide range of purposes. Following that, a modified Mask Regional Convolutional Neural Network (Mask R-CNN) is proposed, consisting of a region-based CNN for finger detection and a three-layer CNN for fingertip location. The process can be completed in 25 ms per frame for 6 RGB images, with an average error of pixels. The speed is high enough to enable real-time “air-writing”, where users are able to write characters in the air to input texts or commands while wearing smart glasses. The characters can be recognized by a ResNet-based CNN from the fingertip trajectories. Experimental results demonstrate the feasibility of this novel methodology.

Mapping the Masked Region of the A1 Domain in Full-Length VWF

Von Willebrand factor (VWF) is the largest protein in plasma and promotes the early hemostatic response following vascular injury. The hemostatic response of VWF relies on the exposure of its A1 domain, which recognizes its receptor GPIbα on the platelet surface. To avoid binding and activating platelets in the absence of injury, the A1 domain is dormant in full-length VWF through a mechanism that is not completely understood. We recently characterized two recombinant A1 fragments (longA1, residues 1238-1472; and shortA1, residues 1261-1472) with differential GPIbα-binding affinities by hydrogen-deuterium exchange (HDX) mass spectrometry. We found that N- and C-terminal residues flanking the A1 domain (residues 1238-1271 and 1459-1472) cooperatively constitute an autoinhibitory module (AIM) that masks a region in A1 and impedes its binding to GPIbα[1]. To determine if the A1 domain in full-length VWF is “masked” in the same manner, we have characterized human plasma-derived VWF using the same HDX approach.On-column peptic fragmentation was performed following HDX from 10 to 10,000 sec. We obtained 946 sequenced peptides covering 96.4% of the VWF sequence with an average redundancy of 6 per residue. Among them, 113 peptides covered the A1 domain, allowing HDX analysis of A1 at high resolution. We found that residues around the 1272-1458 disulfide bond, helix α1, loops α1β2 and β3α2 in the A1 domain, which are protected by the AIM in the longA1 fragment (residues 1238-1472), are deeply protected in full-length VWF, suggesting that AIM in full-length VWF likely masks A1 in a similar pattern. Additionally, sheet β3, helix α3, and loops β4α3 in the A1 domain of full-length VWF reported lower HDX rates than their counterparts in the shortA1 fragment. These regions, along with those protected by AIM, form a contiguous surface that significantly overlaps with the GPIbα-binding interface (Fig. 1). Residues in loop β6α6, helix α6 and helix α1 also reported lower HDX rates, but they are separate from the regions protected by AIM and the GPIbα-binding interface.In summary, our results suggest that AIM and likely other domains in full-length VWF are in direct contact with the A1 domain and regulate the hemostatic properties of VWF.Figure 1: Difference in HDX between isolated VWF A1 domains and full length VWF.1 Deng W, Wang Y, Druzak SA, Healey JF, Syed AK, Lollar P, Li R. A discontinuous autoinhibitory module masks the A1 domain of von Willebrand factor. Journal of thrombosis and haemostasis : JTH . 2017. 10.1111/jth.13775. [Display omitted] DisclosuresNo relevant conflicts of interest to declare.

Recurrent Genetic Variants of Long Non-Coding RNAs (lncRNAs) Independently Associate with Clinical Outcome of Younger Adults with Cytogenetically Normal Acute Myeloid Leukemia

BACKGROUNDAberrant expression of lncRNAs has been shown to independently associate with clinical outcome of younger adult patients (pts) [aged <60 years (y); Papaioannou et al. Haematologica 2017;102:1391-1400] and older pts (aged ≥60 y; Garzon et al. PNAS 2014;111:18679-84) with cytogenetically normal acute myeloid leukemia (CN-AML). However, the prognostic significance of genetic variants (single nucleotide polymorphisms or somatic mutations) in lncRNAs and their association with established prognostic gene mutations have not been studied. METHODSWe performed whole transcriptome profiling (RNA-seq) in 377 younger adults with de novo CN-AML and investigated the associations of recurrent genetic variants located within expressed lncRNAs with clinical outcome and gene mutations. All pts were treated on frontline Cancer and Leukemia Group B (CALGB)/Alliance for Clinical Trials in Oncology protocols. Analysis of 11 gene mutations, reported to be prognostic in AML, was performed by targeted amplicon sequencing (ASXL1, DNMT3A, FLT3 -TKD, IDH2, IDH1, NPM1, RUNX1, TET2, WT1), Sanger sequencing (double CEBPA) and fragment analysis (FLT3 -ITD). RESULTSWe first generated a comprehensive list of the genetic variants that are located in lncRNAs in transcriptomes of younger CN-AML pts. To analyze unequivocally non-coding genetic variants and to avoid ambiguity in their genomic location, all variants that overlapped exons of protein coding genes and those that mapped to segmental duplications or repeat masked regions of the genome were excluded from further analyses. We focused on 207 variants whose frequency differed between our AML cohort and the healthy population, and on 88 variants that were novel. Of the variants tested, 8 showed significant association with at least 1 outcome endpoint [disease-free (DFS), overall (OS) or event-free (EFS) survival], and 3 of them significantly correlated with DFS, OS and EFS. There was no association between complete remission rates and lncRNA variants. The 3 variants which associated with all survival endpoints were: a thymidine (T)-to-cytosine (C) variant in lncRNA RP5-1074L1.4, an adenosine (A)-to-C variant in lncRNA HCG4, and a C-to-T variant in lncRNA HLA-V . The C variant of RP5-1074L1.4 was detected in 182 (64%) of the 284 pts who expressed the lncRNA and associated with longer DFS (P <.001; 5-y DFS, 41% v 27%), OS (P=.04; 5-y OS, 40% v 35%) and EFS (P <.001; 5-y EFS, 36% v 22%). The C variant of HCG4 was found in 41 (18%) of the 234 pts with detectable expression of the lncRNA. Pts harboring the C variant had longer DFS (P=.01; 5-y DFS, 61% v 35%), OS (P=.02; 5-y OS, 58% v 39%) and EFS (P=.04; 5-year EFS, 49% v 29%). Finally, the T variant of HLA-V, which was detected in 58 (26%) of the 227 pts who expressed the transcript, associated with longer DFS (P=.005; 5-y DFS, 53% v 30%), OS (P=.01; 5-y OS, 55% v 33%) and EFS (P=.007; 5-y EFS, 45% v 24%).With regard to associations with prognostic gene mutations, pts with the C variant of RP5-1074L1.4 were more likely to harbor FLT3 -TKD than pts with the T variant (P=.02). Pts with the T variant in HLA-V were less likely to harbor RUNX1 mutations than pts with the C variant (P=.02). The C variant of HCG4 did not associate with differences in the frequencies of prognostic mutations.In multivariable analyses, the variables considered in the models (other than the prognostic lncRNA variants) were: white blood cell and platelet counts, hemoglobin, extramedullary involvement, the presence of FLT3 -TKD, FLT3 -ITD and CEBPA, DNMT3A, NPM1, RUNX1 and WT1 mutations, and expression levels of MN1 and miR-155. The C variant of RP5-1074L1.4 was an independent marker for longer DFS (HR: 0.53, P <.001), OS (HR: 0.64, P=.007) and EFS (HR: 0.55, P <.001). The C variant of HCG4 independently associated with longer DFS (HR: 0.52, P=.03) and OS (HR: 0.61, P=.05), whereas the T variant of HLA-V independently associated with longer DFS (HR: 0.46, P=.003), OS (HR: 0.38, P <.001) and EFS (HR: 0.48, P=.001).PCR assays and targeted Sanger sequencing experiments in leukemic blasts and paired remission bone marrow samples are currently being conducted to examine whether the prognostic variants are acquired genetic events or germline variants. CONCLUSIONSWe conclude that a small subset of genetic variants that are located within lncRNAs are clinically relevant and provide independent prognostic information in younger adults with CN-AML. DisclosuresBlum:Astellas: Consultancy; Pfizer: Consultancy; Boerhinger Ingelheim: Research Funding. Uy:Boehringer Ingelheim: Consultancy; Novartis: Consultancy, Other: Travel Suppport; GlycoMimetics: Consultancy. Stone:Celgene: Consultancy; Novartis: Consultancy; Fuji Film: Consultancy; Ono: Consultancy; Agios: Consultancy; Astellas: Consultancy; Abbvie: Consultancy; Amgen: Consultancy; Jazz: Consultancy; Sumitomo: Consultancy; Pfizer: Consultancy; Arog: Consultancy. Byrd:Acerta Pharma: Research Funding; Janssen: Research Funding; The Ohio State University: Patents & Royalties: OSU-2S; Genentech: Research Funding; Pharmacyclics: Research Funding.

Object Identification and Localization Using Grad-CAM++ with Mask Regional Convolution Neural Network

One of the fundamental advancements in the deployment of object detectors in real-time applications is to improve object recognition against obstruction, obscurity, and noises in images. In addition, object detection is a challenging task since it needs the correct detection of objects from images. Semantic segmentation and localization are an important module to recognizing an object in an image. The object localization method (Grad-CAM++) is mostly used by researchers for object localization, which uses the gradient with a convolution layer to build a localization map for important regions on the image. This paper proposes a method called Combined Grad-CAM++ with the Mask Regional Convolution Neural Network (GC-MRCNN) in order to detect objects in the image and also localization. The major advantage of proposed method is that they outperform all the counterpart methods in the domain and can also be used in unsupervised environments. The proposed detector based on GC-MRCNN provides a robust and feasible ability in detecting and classifying objects exist and their shapes in real time. It is found that the proposed method is able to perform highly effectively and efficiently in a wide range of images and provides higher resolution visual representation than existing methods (Grad-CAM, Grad-CAM++), which was proven by comparing various algorithms.

Evaluation of Multi-Stream Fusion for Multi-View Image Set Comparison

We consider the problem of image set comparison, i.e., to determine whether two image sets show the same unique object (approximately) from the same viewpoints. Our proposition is to solve it by a multi-stream fusion of several image recognition paths. Immediate applications of this method can be found in fraud detection, deduplication procedure, or visual searching. The contribution of this paper is a novel distance measure for similarity of image sets and the experimental evaluation of several streams for the considered problem of same-car image set recognition. To determine a similarity score of image sets (this score expresses the certainty level that both sets represent the same object visible from the same set of views), we adapted a measure commonly applied in blind signal separation (BSS) evaluation. This measure is independent of the number of images in a set and the order of views in it. Separate streams for object classification (where a class represents either a car type or a car model-and-view) and object-to-object similarity evaluation (based on object features obtained alternatively by the convolutional neural network (CNN) or image keypoint descriptors) were designed. A late fusion by a fully-connected neural network (NN) completes the solution. The implementation is of modular structure—for semantic segmentation we use a Mask-RCNN (Mask regions with CNN features) with ResNet 101 as a backbone network; image feature extraction is either based on the DeepRanking neural network or classic keypoint descriptors (e.g., scale-invariant feature transform (SIFT)) and object classification is performed by two Inception V3 deep networks trained for car type-and-view and car model-and-view classification (4 views, 9 car types, and 197 car models are considered). Experiments conducted on the Stanford Cars dataset led to selection of the best system configuration that overperforms a base approach, allowing for a 67.7% GAR (genuine acceptance rate) at 3% FAR (false acceptance rate).

DETECTING CRACKS AND SPALLING AUTOMATICALLY IN EXTREME EVENTS BY END-TO-END DEEP LEARNING FRAMEWORKS

Abstract. In this paper, we develop and implement end-to-end deep learning approaches to automatically detect two important types of structural failures, cracks and spalling, of buildings and bridges in extreme events such as major earthquakes. A total of 2,229 images were annotated, and are used to train and validate three newly developed Mask Regional Convolutional Neural Networks (Mask R-CNNs). In addition, three sets of public images for different disasters were used to test the accuracy of these models. For detecting and marking these two types of structural failures, one of proposed methods can achieve an accuracy of 67.6% and 81.1%, respectively, on low- and high-resolution images collected from field investigations. The results demonstrate that it is feasible to use the proposed end-to-end method for automatically locating and segmenting the damage using 2D images which can help human experts in cases of disasters.

A deep learning-based vision enhancement method for UAV assisted visual inspection of concrete cracks

Implementing unmanned aerial vehicles (UAVs) on concrete surface-crack inspection leads to a promising visual crack detection approach. One of the challenges for automated field visual cracking inspection is image degradation caused by the rain or fog and motion blur during data acquisition. The present study combines two deep neural networks to address the image degradation problem. By using the Variance of Laplacian algorithm for quantifying image clarity, the proposed deep neural networks can remarkably enhance the sharpness of the degraded images. After vision enhancement process, Mask Region Convolutional Neutral Network (Mask R-CNN) was developed to perform automated crack identification and segmentation. Results show a 8~13% enhancement in prediction accuracy compared to the degraded images, indicating that the proposed deep learning-based vision enhancement method can effectivey identify and segment concrete surface cracks from photos captured by UAVs.

Dark Energy Survey Year 3 Results: Photometric Data Set for Cosmology

Abstract We describe the Dark Energy Survey (DES) photometric data set assembled from the first three years of science operations to support DES Year 3 cosmologic analyses, and provide usage notes aimed at the broad astrophysics community. Y3 GOLD improves on previous releases from DES, Y1 GOLD, and Data Release 1 (DES DR1), presenting an expanded and curated data set that incorporates algorithmic developments in image detrending and processing, photometric calibration, and object classification. Y3 GOLD comprises nearly 5000 deg2 of grizY imaging in the south Galactic cap, including nearly 390 million objects, with depth reaching a signal-to-noise ratio ∼10 for extended objects up to i AB ∼ 23.0, and top-of-the-atmosphere photometric uniformity &lt;3 mmag. Compared to DR1, photometric residuals with respect to Gaia are reduced by 50%, and per-object chromatic corrections are introduced. Y3 GOLD augments DES DR1 with simultaneous fits to multi-epoch photometry for more robust galactic color measurements and corresponding photometric redshift estimates. Y3 GOLD features improved morphological star–galaxy classification with efficiency &gt;98% and purity &gt;99% for galaxies with 19 &lt; i AB &lt; 22.5. Additionally, it includes per-object quality information, and accompanying maps of the footprint coverage, masked regions, imaging depth, survey conditions, and astrophysical foregrounds that are used to select the cosmologic analysis samples.

An intelligent identification system combining image and DNA sequence methods for fruit flies with economic importance (Diptera: Tephritidae).

Images and DNA sequences are two important methods for identifying fruit fly species. In addition, the identification of insect species complexes is highly problematic when attempting to utilize automatic identification methods in an actual environment. We integrated the image and DNA sequence identification methods into a single system for the first time and explored an open interactive multi-image comparison function for solving the problem of species complexes. The Automated Fruit Fly Identification System 1.0 (AFIS1.0) was updated to AFIS2.0 by employing different models and developing the system under a novel framework. AFIS2.0 was developed using 83 species belonging to eight genera in the Tephritidae, which includes most pests of this family. The system applies the Mask Region Convolutional Neural Network (Mask R-CNN) and discriminative deep metric learning (AlexNet based) methods for image identification, integrates Blast+ for DNA sequence comparison and specific weighting for the fusion result. At the species level, the best classification success rate for wing images (as the Top 1 species in the species list of outcomes) reached 90%, and the average classification success rate for wing, thorax, and abdomen images (as the Top 5 species in the species list of outcomes) was 94%. AFIS2.0 is more accurate and convenient than AFIS1.0 and can be beneficial for users with or without specific expertise regarding Tephritidae. It also provides a more compact and fluent computer system for fruit fly identification, and can be easily applied in practice. © 2021 Society of Chemical Industry.

Object Identification and Localization using Convolution Neural Network

Improving object identification against impediment, obscure and clamor image is a basic advance to deploy detector in real time applications. Since it is preposterous to expect to debilitate all picture abandons through information assortment, numerous specialists look to produce hard examples in preparing. The produced hard examples are either pictures or highlight maps with coarse patches exited in the spatial measurements. Huge overheads are needed in preparing the extra hard examples and additionally assessing drop-out patches utilizing additional organization branches. In this paper we proposed GRAD CAM++ with Mask Regional Convolution Neural Network (Mask RCNN) based item limitation and identification. The significant advantages of utilizing Mask R-CNN is that they beat all the partner techniques in the space and can likewise be utilized in unaided environments. The proposed identifier dependent on GRAD CAM++ with Mask R-CNN gives a vigorous and plausible capacity on recognizing and grouping objects exist and their shapes progressively on location. It is discovered that the proposed strategy can perform exceptionally successful and productive in a wide scope of pictures and gives higher goal visual portrayal.

Avoiding the Geometric Boundary Effect in Shear Measurement

Abstract In image processing, source detections are inevitably affected by the presence of the geometric boundaries in the images, including the physical boundaries of the CCD, and the boundaries of masked regions due to column defects, bright diffraction spikes, etc. These boundary conditions make the source detection process not statistically isotropic and can lead to additive shear bias near the boundaries. We build a phenomenological model to understand the bias, and propose a simple method to effectively eliminate the influence of geometric boundaries on shear measurement. We demonstrate the accuracy and efficiency of this method using both simulations and the z-band imaging data from the third data release of the DECam Legacy Survey.

A GF-3 SAR Image Dataset of Road Segmentation

We constructed a GF-3 SAR image dataset based on road segmentation to boost the development of GF-3 synthetic aperture radar (SAR) image road segmentation technology and make GF-3 SAR images be applied to practice better. We selected 23 scenes of GF-3 SAR images in Shaanxi, China, cut them into road chips with 512 × 512 pixels, and then labeled the dataset using LabelMe labeling tool. The dataset consists of 10026 road chips, and these road images are from different GF-3 imaging modes, so there is diversity in resolution and polarization. Three segmentation algorithms such as Multi-task Network Cascades (MNC), Fully Convolutional Instance-aware Semantic Segmentation (FCIS), and Mask Region Convolutional Neural Networks (Mask R-CNN) are trained by using the dataset. The experimental result measures including Average Precision (AP) and Intersection over Union (IoU) show that segmentation algorithms work well with this dataset, and the segmentation accuracy of Mask R-CNN is the best, which demonstrates the validity of the dataset we constructed.

Real-time molecular-level visualization of mass flow during patterned photopolymerization of liquid-crystalline monomers

Single-particle fluorescence imaging is used to monitor dynamic processes that occur during patterned photopolymerization of liquid-crystalline monomers. A spatial gradient of chemical potential can be created at the border of bright and dark regions by structured illumination in the photopolymerization process, leading to mutual diffusion of polymers and monomers. Analysis of the fluorescence from single quantum dots doped into the monomers at minute concentrations enables visualization of highly directional flow from the illuminated region where the photopolymerization proceeds toward a masked unpolymerized region. This directional mass flow causes flow-induced orientation of the polymers that is subsequently fixed by completion of the polymerization reaction, resulting in a mesoscopic aligned area of the polymer film.

Mapping trees along urban street networks with deep learning and street-level imagery

Planning and managing urban forests for livable cities remains a challenge worldwide owing to sparse information on the spatial distribution, structure and composition of urban trees and forests. National and municipal sources of tree inventory remain limited due to a lack of detailed, consistent and frequent inventory assessments. Despite advancements in research on the automation of urban tree mapping using Light Detection and Ranging (LiDAR) or high-resolution satellite imagery, in practice most municipalities still perform labor-intensive field surveys to collect and update tree inventories. We present a robust, affordable and rapid method for creating tree inventories in any urban region where sufficient street-level imagery is readily available. Our approach is novel in that we use a Mask Regional Convolutional Neural Network (Mask R-CNN) to detect and locate separate tree instances from street-level imagery, thereby successfully creating shape masks around unique fuzzy urban objects like trees. The novelty of this method is enhanced by using monocular depth estimation and triangulation to estimate precise tree location, relying only on photographs and images taken from the street. Experiments across four cities show that our method is transferable to different image sources (Google Street View, Mapillary) and urban ecosystems. We successfully detect >70% of all public and private trees recorded in a ground-truth campaign across Metro Vancouver. The accuracy of geolocation is also promising. We automatically locate public and private trees with a mean error in the absolute position ranging from 4 to 6 m, which is comparable to ground-truth measurements in conventional manual urban tree inventory campaigns.