Quantification and Predictors of Visual Field Variability in Healthy, Glaucoma Suspect, and Glaucomatous Eyes Using SITA-Faster.

PURPOSE: The newly released Swedish Interactive Thresholding Algorithm (SITA)-Faster (SFR) has significantly shorter testing durations compared with older SITA algorithms, but its variability is uncertain. This study quantified and established threshold limits of test-retest variability across the 24-2 test grid using SFR. DESIGN: Cross-sectional study with prospective longitudinal arm. PARTICIPANTS: 1426 eyes of 787 patients with healthy, suspected glaucoma, or manifest glaucoma eyes from hospital- and university- eye clinics. METHODS: Two SFR tests per eye at a baseline visit and at two follow-up visits. MAIN OUTCOME MEASURES: Pointwise variability measured by test-retest difference in pointwise sensitivity between tests one and two, mean global variability (test-retest variance) measured by average of pointwise variability for each participant, global sensitivity, and reliability indices of each eye. RESULTS: Of the 1426 eyes, 540 eyes (37.9%) had a diagnosis of glaucoma, 753 eyes (52.8%) were suspected of having glaucoma, and the remaining 133 eyes (9.3%) were healthy. Of 74 152 pointwise sensitivities obtained, the mean test-retest difference was 2.17 ± 2.9 dB, whereas the mean test-retest variance for each participant was 2.17 ± 1.2 dB. Pointwise and global variability increased with worsening threshold sensitivity and (MD), respectively, and was greater for peripheral compared with central test locations. In the longitudinal cohort, no significant difference in mean test-retest variance was found across the 3 visits (mean variability, 2.10 dB vs. 2.16 dB vs. 2.16 dB at visits F0 vs. F1 vs. F2; P = 0.53, repeated-measures analysis of variance). Baseline MD (-0.19 dB; 95% CI, -0.22 to 0.16 dB; P < 0.0001) and abnormally high sensitivity on glaucoma hemifield test (1.14 dB; 95% CI, 0.78-1.51 dB; P < 0.0001) were significantly associated with increased variability. Finally, test-retest MD showed minimal change around the recommended 15% false-positive cutoff threshold. CONCLUSIONS: The variability of SFR increases with worsening threshold sensitivity, is stable over time, and is greater for peripheral compared with central test locations. Worse baseline MD and abnormally high sensitivity are significant predictors of increased variability. A cutoff of 15% in false-positive results may be inappropriate as a threshold for judging test reliability in SFR. FINANCIAL DISCLOSURE(S): The author(s) have no proprietary or commercial interest in any materials discussed in this article.

DeepLeish: a deep learning based support system for the detection of Leishmaniasis parasite from Giemsa-stained microscope images.

BACKGROUND: Leishmaniasis is a vector-born neglected parasitic disease belonging to the genus Leishmania. Out of the 30 Leishmania species, 21 species cause human infection that affect the skin and the internal organs. Around, 700,000 to 1,000,000 of the newly infected cases and 26,000 to 65,000 deaths are reported worldwide annually. The disease exhibits three clinical presentations, namely, the cutaneous, muco-cutaneous and visceral Leishmaniasis which affects the skin, mucosal membrane and the internal organs, respectively. The relapsing behavior of the disease limits its diagnosis and treatment efficiency. The common diagnostic approaches follow subjective, error-prone, repetitive processes. Despite, an ever pressing need for an accurate detection of Leishmaniasis, the research conducted so far is scarce. In this regard, the main aim of the current research is to develop an artificial intelligence based detection tool for the Leishmaniasis from the Geimsa-stained microscopic images using deep learning method. METHODS: Stained microscopic images were acquired locally and labeled by experts. The images were augmented using different methods to prevent overfitting and improve the generalizability of the system. Fine-tuned Faster RCNN, SSD, and YOLOV5 models were used for object detection. Mean average precision (MAP), precision, and Recall were calculated to evaluate and compare the performance of the models. RESULTS: The fine-tuned YOLOV5 outperformed the other models such as Faster RCNN and SSD, with the MAP scores, of 73%, 54% and 57%, respectively. CONCLUSION: The currently developed YOLOV5 model can be tested in the clinics to assist the laboratorists in diagnosing Leishmaniasis from the microscopic images. Particularly, in low-resourced healthcare facilities, with fewer qualified medical professionals or hematologists, our AI support system can assist in reducing the diagnosing time, workload, and misdiagnosis. Furthermore, the dataset collected by us will be shared with other researchers who seek to improve upon the detection system of the parasite. The current model detects the parasites even in the presence of the monocyte cells, but sometimes, the accuracy decreases due to the differences in the sizes of the parasite cells alongside the blood cells. The incorporation of cascaded networks in future and the quantification of the parasite load, shall overcome the limitations of the currently developed system.

Enhancing assisted diagnostic accuracy in scalp psoriasis: A Multi-Network Fusion Object Detection Framework for dermoscopic pattern diagnosis.

BACKGROUND: Dermoscopy is a common method of scalp psoriasis diagnosis, and several artificial intelligence techniques have been used to assist dermoscopy in the diagnosis of nail fungus disease, the most commonly used being the convolutional neural network algorithm; however, convolutional neural networks are only the most basic algorithm, and the use of object detection algorithms to assist dermoscopy in the diagnosis of scalp psoriasis has not been reported. OBJECTIVES: Establishment of a dermoscopic modality diagnostic framework for scalp psoriasis based on object detection technology and image enhancement to improve diagnostic efficiency and accuracy. METHODS: We analyzed the dermoscopic patterns of scalp psoriasis diagnosed at 72nd Group army hospital of PLA from January 1, 2020 to December 31, 2021, and selected scalp seborrheic dermatitis as a control group. Based on dermoscopic images and major dermoscopic patterns of scalp psoriasis and scalp seborrheic dermatitis, we investigated a multi-network fusion object detection framework based on the object detection technique Faster R-CNN and the image enhancement technique contrast limited adaptive histogram equalization (CLAHE), for assisting in the diagnosis of scalp psoriasis and scalp seborrheic dermatitis, as well as to differentiate the major dermoscopic patterns of the two diseases. The diagnostic performance of the multi-network fusion object detection framework was compared with that between dermatologists. RESULTS: A total of 1876 dermoscopic images were collected, including 1218 for scalp psoriasis versus 658 for scalp seborrheic dermatitis. Based on these images, training and testing are performed using a multi-network fusion object detection framework. The results showed that the test accuracy, specificity, sensitivity, and Youden index for the diagnosis of scalp psoriasis was: 91.0%, 89.5%, 91.0%, and 0.805, and for the main dermoscopic patterns of scalp psoriasis and scalp seborrheic dermatitis, the diagnostic results were: 89.9%, 97.7%, 89.9%, and 0.876. Comparing the diagnostic results with those of five dermatologists, the fusion framework performs better than the dermatologists' diagnoses. CONCLUSIONS: Studies have shown some differences in dermoscopic patterns between scalp psoriasis and scalp seborrheic dermatitis. The proposed multi-network fusion object detection framework has higher diagnostic performance for scalp psoriasis than for dermatologists.

Comparing a head-mounted virtual reality perimeter and the Humphrey Field Analyzer for visual field testing in healthy and glaucoma patients.

PURPOSE: To compare clinical visual field outputs in glaucoma and healthy patients returned by the Humphrey Field Analyzer (HFA) and virtual reality (Virtual Field, VF) perimetry. METHODS: One eye of 54 glaucoma patients and 41 healthy subjects was prospectively tested (three times each in random order) using the HFA and VF perimeters (24-2 test grids). We extracted and compared global indices (mean deviation [MD] and pattern standard deviation [PSD]), pointwise sensitivity (and calculated 'equivalent' sensitivity after accounting for differences in background luminance) and pointwise defects. Bland-Altman (mean difference [Mdiff ] and 95% limits of agreement [LoA]) and intraclass correlation analyses were performed. RESULTS: The VF test was shorter (by 76 s) and had lower fixation losses (by 0.08) and false-positive rate (by 0.01) compared to the HFA (all p < 0.0001). Intraclass correlations were 0.86, 0.82 and 0.47 for MD, PSD and pointwise sensitivity between devices, respectively. Test-retest variability was higher for VF (Mdiff 0.3 dB, LoA -7.6 to 8.2 dB) compared to the HFA (Mdiff -0.3 dB, LoA -6.4 to 5.9 dB), indicating greater test-retest variability. When using each device's underlying normative database, the HFA detected, on average, 7 more defects (at the p < 0.05 level) out of the 52 test locations compared to this iteration of VF in the glaucoma cohort. CONCLUSIONS: Virtual Field returns global results that are correlated with the HFA, but pointwise sensitivities were more variable. Differences in test-retest variability and defect detection by its current normative database raise questions about the widespread adoption of VF in lieu of the HFA.

Frontloading visual field tests detect earlier mean deviation progression when applied to real-world-derived early-stage glaucoma data.

PURPOSE: To examine the diagnostic accuracy of performing two (frontloaded) versus one (clinical standard) visual field (VF) test per visit for detecting the progression of early glaucoma in data derived from clinical populations. METHODS: A computer simulation model was used to follow the VFs of 10,000 glaucoma patients (derived from two cohorts: Heijl et al., Swedish cohort; and Chauhan et al., Canadian Glaucoma Study [CGS]) over a 10-year period to identify patients whose mean deviation (MD) progression was detected. Core data (baseline MD and progression rates) were extracted from two studies in clinical cohorts of glaucoma, which were modulated using SITA-Faster variability characteristics from previous work. Additional variables included follow-up intervals (six-monthly or yearly) and rates of perimetric data loss for any reason (0%, 15% and 30%). The main outcome measures were the proportions of progressors detected. RESULTS: When the Swedish cohort was reviewed six-monthly, the frontloaded strategy detected more progressors compared to the non-frontloaded method up to years 8, 9 and 10 of follow-up for 0%, 15% and 30% data loss conditions. The time required to detect 50% of cases was 1.0-1.5 years less for frontloading compared to non-frontloading. At 4 years, frontloading increased detection by 26.7%, 28.7% and 32.4% for 0%, 15% and 30% data loss conditions, respectively. Where both techniques detected progression, frontloading detected progressors earlier compared to the non-frontloaded strategy (78.5%-81.5% and by 1.0-1.3 years when reviewed six-monthly; 81%-82.9% and by 1.2-2.1 years when reviewed yearly). Accordingly, these patients had less severe MD scores (six-monthly review: 0.63-1.67 dB 'saved'; yearly review: 1.10-2.87 dB). The differences increased with higher rates of data loss. Similar tendencies were noted when applied to the CGS cohort. CONCLUSIONS: Frontloaded VFs applied to clinical distributions of MD and progression led to earlier detection of early glaucoma progression.

Deep Learning-Based Fish Detection Using Above-Water Infrared Camera for Deep-Sea Aquaculture: A Comparison Study.

Long-term, automated fish detection provides invaluable data for deep-sea aquaculture, which is crucial for safe and efficient seawater aquafarming. In this paper, we used an infrared camera installed on a deep-sea truss-structure net cage to collect fish images, which were subsequently labeled to establish a fish dataset. Comparison experiments with our dataset based on Faster R-CNN as the basic objection detection framework were conducted to explore how different backbone networks and network improvement modules influenced fish detection performances. Furthermore, we also experimented with the effects of different learning rates, feature extraction layers, and data augmentation strategies. Our results showed that Faster R-CNN with the EfficientNetB0 backbone and FPN module was the most competitive fish detection network for our dataset, since it took a significantly shorter detection time while maintaining a high AP50 value of 0.85, compared to the best AP50 value of 0.86 being achieved by the combination of VGG16 with all improvement modules plus data augmentation. Overall, this work has verified the effectiveness of deep learning-based object detection methods and provided insights into subsequent network improvements.

Vehicle-Type Recognition Method for Images Based on Improved Faster R-CNN Model.

The rapid increase in the number of vehicles has led to increasing traffic congestion, traffic accidents, and motor vehicle crime rates. The management of various parking lots has also become increasingly challenging. Vehicle-type recognition technology can reduce the workload of humans in vehicle management operations. Therefore, the application of image technology for vehicle-type recognition is of great significance for integrated traffic management. In this paper, an improved faster region with convolutional neural network features (Faster R-CNN) model was proposed for vehicle-type recognition. Firstly, the output features of different convolution layers were combined to improve the recognition accuracy. Then, the average precision (AP) of the recognition model was improved through the contextual features of the original image and the object bounding box optimization strategy. Finally, the comparison experiment used the vehicle image dataset of three vehicle types, including cars, sports utility vehicles (SUVs), and vans. The experimental results show that the improved recognition model can effectively identify vehicle types in the images. The AP of the three vehicle types is 83.2%, 79.2%, and 78.4%, respectively, and the mean average precision (mAP) is 1.7% higher than that of the traditional Faster R-CNN model.

Apple Fruit Edge Detection Model Using a Rough Set and Convolutional Neural Network.

Accurately and effectively detecting the growth position and contour size of apple fruits is crucial for achieving intelligent picking and yield predictions. Thus, an effective fruit edge detection algorithm is necessary. In this study, a fusion edge detection model (RED) based on a convolutional neural network and rough sets was proposed. The Faster-RCNN was used to segment multiple apple images into a single apple image for edge detection, greatly reducing the surrounding noise of the target. Moreover, the K-means clustering algorithm was used to segment the target of a single apple image for further noise reduction. Considering the influence of illumination, complex backgrounds and dense occlusions, rough set was applied to obtain the edge image of the target for the upper and lower approximation images, and the results were compared with those of relevant algorithms in this field. The experimental results showed that the RED model in this paper had high accuracy and robustness, and its detection accuracy and stability were significantly improved compared to those of traditional operators, especially under the influence of illumination and complex backgrounds. The RED model is expected to provide a promising basis for intelligent fruit picking and yield prediction.

Molecular Evolution across Mouse Spermatogenesis.

Genes involved in spermatogenesis tend to evolve rapidly, but we lack a clear understanding of how protein sequences and patterns of gene expression evolve across this complex developmental process. We used fluorescence-activated cell sorting (FACS) to generate expression data for early (meiotic) and late (postmeiotic) cell types across 13 inbred strains of mice (Mus) spanning ∼7 My of evolution. We used these comparative developmental data to investigate the evolution of lineage-specific expression, protein-coding sequences, and expression levels. We found increased lineage specificity and more rapid protein-coding and expression divergence during late spermatogenesis, suggesting that signatures of rapid testis molecular evolution are punctuated across sperm development. Despite strong overall developmental parallels in these components of molecular evolution, protein and expression divergences were only weakly correlated across genes. We detected more rapid protein evolution on the X chromosome relative to the autosomes, whereas X-linked gene expression tended to be relatively more conserved likely reflecting chromosome-specific regulatory constraints. Using allele-specific FACS expression data from crosses between four strains, we found that the relative contributions of different regulatory mechanisms also differed between cell types. Genes showing cis-regulatory changes were more common late in spermatogenesis, and tended to be associated with larger differences in expression levels and greater expression divergence between species. In contrast, genes with trans-acting changes were more common early and tended to be more conserved across species. Our findings advance understanding of gene evolution across spermatogenesis and underscore the fundamental importance of developmental context in molecular evolutionary studies.

A Test of Haldane's Rule in Neodiprion Sawflies and Implications for the Evolution of Postzygotic Isolation in Haplodiploids.

AbstractHaldane's rule-a pattern in which hybrid sterility or inviability is observed in the heterogametic sex of an interspecific cross-is one of the most widely obeyed rules in nature. Because inheritance patterns are similar for sex chromosomes and haplodiploid genomes, Haldane's rule may apply to haplodiploid taxa, predicting that haploid male hybrids will evolve sterility or inviability before diploid female hybrids. However, there are several genetic and evolutionary mechanisms that may reduce the tendency of haplodiploids to obey Haldane's rule. Currently, there are insufficient data from haplodiploids to determine how frequently they adhere to Haldane's rule. To help fill this gap, we crossed a pair of haplodiploid hymenopteran species (Neodiprion lecontei and Neodiprion pinetum) and evaluated the viability and fertility of female and male hybrids. Despite considerable divergence, we found no evidence of reduced fertility in hybrids of either sex, consistent with the hypothesis that hybrid sterility evolves slowly in haplodiploids. For viability, we found a pattern opposite to that of Haldane's rule: hybrid females, but not males, had reduced viability. This reduction was most pronounced in one direction of the cross, possibly due to a cytoplasmic-nuclear incompatibility. We also found evidence of extrinsic postzygotic isolation in hybrids of both sexes, raising the possibility that this form or reproductive isolation tends to emerge early in speciation in host-specialized insects. Our work emphasizes the need for more studies on reproductive isolation in haplodiploids, which are abundant in nature but underrepresented in the speciation literature.

Nucleation-Supersaturation Dual-Drive Crystallization Strategy Enables Efficient Protein Crystallization.

A rational crystallization strategy is essential to obtain high-quality protein crystals, yet the established methods suffer from different limitations arising from the single regulation on either nucleation or supersaturation. Herein, a nucleation-supersaturation dual-driven crystallization (DDC) strategy that realizes synergistic regulation of heterogeneous nucleation sites and solution supersaturation based on dual surface and confinement effects for efficient protein crystallization is reported. This strategy relies on a p(PEGDA-co-DMAA) hydrogel template with pre-filled NaCl under designed concentrations. Once dropping hen egg white lysozyme (HEWL) protein solution on the hydrogel, the wrinkled surface provides numerous nucleation sites, while the internal structure regulates the solution supersaturation in the crystallization region through diffusion. Finally, DDC strategy can create high-quality HEWL crystals with large sizes (100-300 µm), well-defined morphologies (hexagon and tetragon), and a significantly accelerated nucleation time (9-12 times faster than that achieved using the conventional hanging drop method). It also performs well at wider protein concentrations (10-50 mg mL-1 ) and categories (e.g., achieving fast crystallization and large-size crystals of trypsin), therefore demonstrating clear advantages and great potential for efficiently fabricating protein crystals desirable for diverse applications.

Lower Diffusion-Induced Stress in Nano-Crystallites of P2-Na2/3 Ni1/3 Mn1/2 Ti1/6 O2 Novel Cathode for High Energy Na-ion Batteries.

P2-type Na2/3 Ni1/3 Mn1/2 Ti1/6 O2 (NMTNO) cathode is a preeminent electrode material for Na-ion batteries owing to its open prismatic framework, air-moisture stability, inexpensiveness, appealing capacity, environmental benignity, and Co-free composition. However, the poor cycling stability, sluggish Na-ion kinetics induced in bulk-sized cathode particles, cracking, and exfoliation in the crystallites remain a setback. To outmaneuver these, a designing strategy of a mechanically robust, hexagonal nano-crystallites of P2-type Na2/3 Ni1/3 Mn1/2 Ti1/6 O2 (NMTNOnano ) electrode via quick, energy-efficient, and low-cost microwave-irradiated synthesis is proposed. For the first time, employing a unified experimental and theoretical approach with fracture mechanics analysis, the mechanism behind the enhanced performance, better structural stability, and lower diffusion-induced stress of NMTNOnano compared to micro-sized Na2/3 Ni1/3 Mn1/2 Ti1/6 O2 is unveiled and the electrochemical shock map is predicted. The NMTNOnano cathode provides 94.8% capacity retention after 100 cycles at 0.1 C with prolonged performance for 1000 cycles at 0.5 C. The practical viability of this cathode, tested in a full cell against a hard carbon anode delivered 85.48% capacity retention at 0.14 mA cm-2 after 200 cycles. This work bridges the gap in correlating the microstructural and electrochemical properties through experimental, theoretical (DFT), and fracture mechanics analysis, thereby tailoring efficient cathode with lower diffusion-induced stress for high-energy Na-ion batteries.

Evaluation of the Consistency of Glaucomatous Visual Field Defects Using a Clustered SITA-Faster Protocol.

PURPOSE: Frontloading SITA-Faster (SFR) visual fields (2 tests per eye on the same visit) has been shown to provide repeatable perimetric data at minimal time cost. This study reports the outcomes of using frontloaded SFR in the evaluation of pointwise visual field (VF) defects in a cohort of patients with glaucoma when transitioned from SITA-Standard (SS). DESIGN: Prospective, cross-sectional study. PARTICIPANTS: A total of 144 eyes of 91 patients with confirmed or suspected glaucoma who had an SS test on a previous visit. METHODS: Two SFR tests (T1, T2) per eye on the same visit. MAIN OUTCOME MEASURES: Global sensitivity, reliability indices, and pointwise deviation map probability scores from the pattern deviation grid of each patient were compared across the 3 sequential tests to evaluate the consistency of VF defects. RESULTS: The mean age was 68.6 years, and 79.2% of patients had a diagnosis of glaucoma. There was no significant difference in mean deviation (MD) across the 3 tests (-5.83 decibels [dB], -5.28 dB, and -5.71 dB in SS, SFR1, and SFR2, respectively, repeated-measures analysis of variance [ANOVA], P = 0.48). The frontloaded SFR tests provided repeatable VFs that confirmed existing pointwise data on the SS in 4661 (62.3%) locations, reversed an SS defect in 614 (8.2%) locations, and demonstrated a new repeatable defect in 406 (5.4%) locations of the pattern deviation grid. A new defect of at least 3 contiguous points was identified in 20.1% of eyes. The non-repeatable points on the 2 SFR tests displayed no significant difference in the distribution of defect/nondefect points based on test order or peripheral versus central locations. There was no significant difference in the rate of obtaining at least 1 reliable test result between SS and the frontloaded SFR T1 and T2 (P = 0.77). Test duration significantly decreased from SS to SFR1/2 (379 vs. 160 vs. 158 seconds, P < 0.0001). CONCLUSIONS: Frontloading SFR tests can provide repeatable data for the evaluation of the consistency of pattern deviation defects in glaucoma, with no observable decline in performance from test fatigue. This is achieved at equivalent duration and reliability as a single SS test. Frontloading SFR may be helpful in increasing testing frequency/quantity to meet recommended guidelines for progression analysis. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

100 years of Haldane's rule.

Haldane's rule is one of the 'two rules of speciation'. It states that if one sex is 'absent, rare or sterile' in a hybrid population, then that sex will be heterogametic. Since Haldane first made this observation, 100 years have passed and still questions arise over how many independent examples exist and what the underlying causes of Haldane's rule are. This review aims to examine research that has occurred over the last century. It seeks to do so by discussing possible causes of Haldane's rule, as well as gaps in the research of these causes that could be readily addressed today. After 100 years of research, it can be concluded that Haldane's rule is a complicated one, and much current knowledge has been accrued by studying the model organisms of speciation. This has led to the primacy of dominance theory and faster-male theory as explanations for Haldane's rule. However, some of the most interesting findings of the 21st century with regard to Haldane's rule have involved investigating a wider range of taxa emphasizing the need to continue using comparative methods, including ever more taxa as new cases are discovered.

Robot-assisted endoscopic submucosal dissection contributes to efficient and safe learning for novices: Prospective pilot cross-over ex vivo study (with video).

OBJECTIVES: The lack of effective countertraction to expose the submucosal layer contributes to the technical complication and adverse events in endoscopic submucosal dissection (ESD). We aimed to evaluate the efficacy and safety of a novel endoscopic robot (flexible auxiliary single-arm transluminal endoscopic robot [FASTER]) for ESD learning for novices. METHODS: This was a prospective, cross-over designed pilot study in ex vivo porcine stomach. Four ESD novices were randomized to either FASTER-assisted ESD first (FC) group or a conventional ESD first (CF) group, performed 40 gastric ESDs using each technique, then crossed over to another technique. The performance and learning curve were compared between the two groups. RESULTS: In the first phase, novices in the FC group demonstrated significantly better performance with shorter procedure time (25.6 ± 7.8 vs. 38.9 ± 13.4 min; P < 0.001) and submucosal dissection time (13.9 ± 5.5 vs. 23.1 ± 11.0 min; P < 0.001), higher direct-vision dissection ratio (84.0 ± 7.9% vs. 43.5 ± 20.7%; P < 0.001), and lower muscular injury (2.5 vs. 40.0%; P < 0.001) and task load (4 vs. 5; P < 0.001). Fewer ESDs were required to gain early proficiency in the FC group. When crossed to the second phase, procedure time in the FC group was prolonged but the muscular injury rate did not increase significantly. In total, endoscopists in the FC group tended to have a lower task load (4 vs. 5; P = 0.008) and less muscular injury (10.0 vs. 21.3%; P = 0.05). CONCLUSION: Flexible auxiliary single-arm transluminal endoscopic robot-assisted learning reduces the technical difficulty of ESD for novices and the safety profile can sustain in following conventional ESD. These results indicated that FASTER has potential implications for ESD training in clinical practice.

Detection and Recognition of Tilted Characters on Railroad Wagon Wheelsets Based on Deep Learning.

The quality of railroad wheelsets is an important guarantee for the safe operation of wagons, and mastering the production information of wheelsets plays a vital role in vehicle scheduling and railroad transportation safety. However, when using objection detection methods to detect the production information of wheelsets, there are situations that affect detection such as character tilting and unfixed position. Therefore, this paper proposes a deep learning-based method for accurately detecting and recognizing tilted character information on railroad wagon wheelsets. It covers three parts. Firstly, we construct a tilted character detection network based on Faster RCNN for generating a wheelset's character candidate regions. Secondly, we design a tilted character correction network to classify and correct the orientation of flipped characters. Finally, a character recognition network is constructed based on convolutional recurrent neural network (CRNN) to realize the task of recognizing a wheelset's characters. The result shows that the method can quickly and effectively detect and identify the information of tilted characters on wheelsets in images.

A Small Object Detection Method for Oil Leakage Defects in Substations Based on Improved Faster-RCNN.

Since substations are key parts of power transmission, ensuring the safety of substations involves monitoring whether the substation equipment is in a normal state. Oil leakage detection is one of the necessary daily tasks of substation inspection robots, which can immediately find out whether there is oil leakage in the equipment in operation so as to ensure the service life of the equipment and maintain the safe and stable operation of the system. At present, there are still some challenges in oil leakage detection in substation equipment: there is a lack of a more accurate method of detecting oil leakage in small objects, and there is no combination of intelligent inspection robots to assist substation inspection workers in judging oil leakage accidents. To address these issues, this paper proposes a small object detection method for oil leakage defects in substations. This paper proposes a small object detection method for oil leakage defects in substations, which is based on the feature extraction network Resnet-101 of the Faster-RCNN model for improvement. In order to decrease the loss of information in the original image, especially for small objects, this method is developed by canceling the downsampling operation and replacing the large convolutional kernel with a small convolutional kernel. In addition, the method proposed in this paper is combined with an intelligent inspection robot, and an oil leakage decision-making scheme is designed, which can provide substation equipment oil leakage maintenance recommendations for substation workers to deal with oil leakage accidents. Finally, the experimental validation of real substation oil leakage image collection is carried out by the intelligent inspection robot equipped with a camera. The experimental results show that the proposed FRRNet101-c model in this paper has the best performance for oil leakage detection in substation equipment compared with several baseline models, improving the Mean Average Precision (mAP) by 6.3%, especially in detecting small objects, which has improved by 12%.

A Deep Learning Approach for Arabic Manuscripts Classification.

For centuries, libraries worldwide have preserved ancient manuscripts due to their immense historical and cultural value. However, over time, both natural and human-made factors have led to the degradation of many ancient Arabic manuscripts, causing the loss of significant information, such as authorship, titles, or subjects, rendering them as unknown manuscripts. Although catalog cards attached to these manuscripts might contain some of the missing details, these cards have degraded significantly in quality over the decades within libraries. This paper presents a framework for identifying these unknown ancient Arabic manuscripts by processing the catalog cards associated with them. Given the challenges posed by the degradation of these cards, simple optical character recognition (OCR) is often insufficient. The proposed framework uses deep learning architecture to identify unknown manuscripts within a collection of ancient Arabic documents. This involves locating, extracting, and classifying the text from these catalog cards, along with implementing processes for region-of-interest identification, rotation correction, feature extraction, and classification. The results demonstrate the effectiveness of the proposed method, achieving an accuracy rate of 92.5%, compared to 83.5% with classical image classification and 81.5% with OCR alone.

Emergency Floor Plan Digitization Using Machine Learning.

An increasing number of special-use and high-rise buildings have presented challenges for efficient evacuations, particularly in fire emergencies. At the same time, however, the use of autonomous vehicles within indoor environments has received only limited attention for emergency scenarios. To address these issues, we developed a method that classifies emergency symbols and determines their location on emergency floor plans. The method incorporates color filtering, clustering and object detection techniques to extract walls, which were used in combination to generate clean, digitized plans. By integrating the geometric and semantic data digitized with our method, existing building information modeling (BIM) based evacuation tools can be enhanced, improving their capabilities for path planning and decision making. We collected a dataset of 403 German emergency floor plans and created a synthetic dataset comprising 5000 plans. Both datasets were used to train two distinct faster region-based convolutional neural networks (Faster R-CNNs). The models were evaluated and compared using 83 floor plan images. The results show that the synthetic model outperformed the standard model for rare symbols, correctly identifying symbol classes that were not detected by the standard model. The presented framework offers a valuable tool for digitizing emergency floor plans and enhancing digital evacuation applications.

A New Assistance Navigation Method for Substation Inspection Robots to Safely Cross Grass Areas.

With the development of intelligent substations, inspection robots are widely used to ensure the safe and stable operation of substations. Due to the prevalence of grass around the substation in the external environment, the inspection robot will be affected by grass when performing the inspection task, which can easily lead to the interruption of the inspection task. At present, inspection robots based on LiDAR sensors regard grass as hard obstacles such as stones, resulting in interruption of inspection tasks and decreased inspection efficiency. Moreover, there are inaccurate multiple object-detection boxes in grass recognition. To address these issues, this paper proposes a new assistance navigation method for substation inspection robots to cross grass areas safely. First, an assistant navigation algorithm is designed to enable the substation inspection robot to recognize grass and to cross the grass obstacles on the route of movement to continue the inspection work. Second, a three-layer convolutional structure of the Faster-RCNN network in the assistant navigation algorithm is improved instead of the original full connection structure for optimizing the object-detection boxes. Finally, compared with several Faster-RCNN networks with different convolutional kernel dimensions, the experimental results show that at the convolutional kernel dimension of 1024, the proposed method in this paper improves the mAP by 4.13% and the mAP is 91.25% at IoU threshold 0.5 in the range of IoU thresholds from 0.5 to 0.9 with respect to the basic network. In addition, the assistant navigation algorithm designed in this paper fuses the ultrasonic radar signals with the object recognition results and then performs the safety judgment to make the inspection robot safely cross the grass area, which improves the inspection efficiency.