Performance Of Deep Convolutional Neural Networks Research Articles

Automated classification of liver fibrosis stages using ultrasound imaging.

Ultrasound imaging is the most frequently performed for the patients with chronic hepatitis or liver cirrhosis. However, ultrasound imaging is highly operator dependent and interpretation of ultrasound images is subjective, thus well-trained radiologist is required for evaluation. Automated classification of liver fibrosis could alleviate the shortage of skilled radiologist especially in low-to-middle income countries. The purposed of this study is to evaluate deep convolutional neural networks (DCNNs) for classifying the degree of liver fibrosis according to the METAVIR score using US images. We used ultrasound (US) images from two tertiary university hospitals. A total of 7920 US images from 933 patients were used for training/validation of DCNNs. All patient were underwent liver biopsy or hepatectomy, and liver fibrosis was categorized based on pathology results using the METAVIR score. Five well-established DCNNs (VGGNet, ResNet, DenseNet, EfficientNet and ViT) was implemented to predict the METAVIR score. The performance of DCNNs for five-level (F0/F1/F2/F3/F4) classification was evaluated through area under the receiver operating characteristic curve (AUC) with 95% confidential interval, accuracy, sensitivity, specificity, positive and negative likelihood ratio. Similar mean AUC values were achieved for five models; VGGNet (0.96), ResNet (0.96), DenseNet (0.95), EfficientNet (0.96), and ViT (0.95). The same mean accuracy (0.94) and specificity values (0.96) were yielded for all models. In terms of sensitivity, EffcientNet achieved highest mean value (0.85) while the other models produced slightly lower values range from 0.82 to 0.84. In this study, we demonstrated that DCNNs can classify the staging of liver fibrosis according to METAVIR score with high performance using conventional B-mode images. Among them, EfficientNET that have fewer parameters and computation cost produced highest performance. From the results, we believe that DCNNs based classification of liver fibrosis may allow fast and accurate diagnosis of liver fibrosis without needs of additional equipment for add-on test and may be powerful tool for supporting radiologists in clinical practice.

Self-Adaptive-Filling Deep Convolutional Neural Network Classification Method for Mountain Vegetation Type Based on High Spatial Resolution Aerial Images

The composition and structure of mountain vegetation are complex and changeable, and thus urgently require the integration of Object-Based Image Analysis (OBIA) and Deep Convolutional Neural Networks (DCNNs). However, while integration technology studies are continuing to increase, there have been few studies that have carried out the classification of mountain vegetation by combining OBIA and DCNNs, for it is difficult to obtain enough samples to trigger the potential of DCNNs for mountain vegetation type classification, especially using high-spatial-resolution remote sensing images. To address this issue, we propose a self-adaptive-filling method (SAF) to incorporate the OBIA method to improve the performance of DCNNs in mountain vegetation type classification using high-spatial-resolution aerial images. Using this method, SAF technology was employed to produce enough regular sample data for DCNNs by filling the irregular objects created by image segmenting using interior adaptive pixel blocks. Meanwhile, non-sample segmented image objects were shaped into different regular rectangular blocks via SAF. Then, the classification result was defined by voting combining the DCNN performance. Compared to traditional OBIA methods, SAF generates more samples for the DCNN and fully utilizes every single pixel of the DCNN input. We design experiments to compare them with traditional OBIA and semantic segmentation methods, such as U-net, MACU-net, and SegNeXt. The results show that our SAF-DCNN outperforms traditional OBIA in terms of accuracy and it is similar to the accuracy of the best performing method in semantic segmentation. However, it reduces the common pretzel phenomenon of semantic segmentation (black and white noise generated in classification). Overall, the SAF-based OBIA using DCNNs, which is proposed in this paper, is superior to other commonly used methods for vegetation classification in mountainous areas.

Ultrasound identification of hepatic echinococcosis using a deep convolutional neural network model in China: a retrospective, large-scale, multicentre, diagnostic accuracy study

Ultrasonography is the most widely used technique to diagnose echinococcosis; however, challenges in using this technique and the demand on medical resources, especially in low-income or remote areas, can delay diagnosis. We aimed to develop a deep convolutional neural network (DCNN) model based on ultrasonography to identify echinococcosis and its types, especially alveolar echinococcosis. This retrospective, large-scale, multicentre study used ultrasound images from patients assessed at 84 hospitals in China, obtained between Jan 1, 2002, and Dec 31, 2021. Patients with a diagnosis of cystic echinococcosis, alveolar echinococcosis, or seven other types of focal liver lesions were included. We tested ResNet-50, ResNext-50, and VGG-16 as the backbone network architecture for a classification DCNN model and input the perinodular information from the ultrasound images. We trained and validated the DCNN model to diagnose and classify echinococcosis using still greyscale ultrasound images of focal liver lesions in four stages: differentiating between echinococcosis and other focal liver lesions (stage one); differentiating cystic echinococcosis, alveolar echinococcosis, and other focal liver lesions (stage two); differentiating cystic echinococcosis, alveolar echinococcosis, benign other focal liver lesions, and malignant focal liver lesions (stage three); and differentiating between active and transitional cystic echinococcosis and inactive cystic echinococcosis (stage four). We then tested the algorithm on internal, external, and prospective test datasets. The performance of DCNN was also compared with that of 12 radiologists recruited between Jan 15, 2022, and Jan 28, 2022, from Qinghai, Xinjiang, Anhui, Henan, Xizang, and Beijing, China, with different levels of diagnostic experience for echinococcosis and other focal liver lesions in a subset of ultrasound data that were randomly chosen from the prospective test dataset. The study is registered at ClinicalTrials.gov (NCT03871140). The study took place between Jan 1, 2002, and Dec 31, 2021. In total, to train and test the DCNN model, we used 9631 liver ultrasound images from 6784 patients (2819 [41·7%] female patients and 3943 [58·3%] male patients) from 87 Chinese hospitals. The DCNN model was trained with 6328 images, internally validated with 984 images, and tested with 2319 images. The ResNet-50 network architecture outperformed VGG-16 and ResNext-50 and was generalisable, with areas under the receiver operating characteristic curve (AUCs) of 0·982 (95% CI 0·960-0·994), 0·984 (0·972-0·992), and 0·913 (0·886-0·935) in distinguishing echinococcosis from other focal liver lesions; 0·986 (0·966-0·996), 0·962 (0·946-0·975), and 0·900 (0·872-0·924) in distinguishing alveolar echinococcosis from cystic echinococcosis and other focal liver lesions; and 0·974 (0·818-1·000), 0·956 (0·875-0·991), and 0·944 (0·844-0·988) in distinguishing active and transitional cystic echinococcosis from inactive echinococcosis in the three test datasets. Specifically, in patients with the hepatitis B or hepatitis C virus, the model could distinguish alveolar echinococcosis from hepatocellular carcinoma with an AUC of 0·892 (0·812-0·946). In identifying echinococcosis, the model showed significantly better performance compared with senior radiologists from a high-endemicity area (AUC 0·942 [0·904-0·967] vs 0·844 [0·820-0·866]; p=0·027) and improved the diagnostic ability of junior, attending, and senior radiologists before and after assistance with AI with comparison of AUCs of 0·743 (0·714-0·770) versus 0·850 (0·826-0·871); p<0·0001, 0·808 (0·782-0·832) versus 0·886 (0·864-0·905); p<0·0001, and 0·844 (0·820-0·866) versus 0·870 (0·847-0·890); p=0·092, respectively. The DCNN model was shown to be accurate and robust, and could improve the ultrasound diagnostic ability of radiologists for echinococcosis and its types for highly endemic and remote regions. National Natural Science Foundation of China and National Key Research & Development Program of China. For the Chinese translation of the abstract see Supplementary Materials section.

Parameter-Free Similarity-Aware Attention Module for Medical Image Classification and Segmentation

Automatic classification and segmentation of medical images play essential roles in computer-aided diagnosis. Deep convolutional neural networks (DCNNs) have shown their advantages on image classification and segmentation. However, they have not achieved the same success on medical images as they have done on natural images. In this paper, two challenges are exploited for DCNNs on medical images, including 1) lack of feature diversity; 2) neglect of small lesions. These two issues heavily influence the classification and segmentation performances. To improve the performance of DCNN on medical images, similarity-aware attention (simi-attention) module is proposed, including a <i>Feature-similarity-aware Channel Attention</i> (FCA) and a <i>Region-similarity-aware Spatial Attention</i> (RSA). Our simi-attention provides three advantages: 1) higher accuracy can be achieved since it extracts both diverse and discriminant features from medical images via our FCA and RSA; 2) the lesions can be exactly focused and located by it even for the data with low intensity contrast and small lesions; 3) it does not increase the complexity of backbone models due to NO trainable parameters in its module. The experimental results are conducted on both classification and segmentation tasks under four public medical classification datasets and two public medical segmentation datasets. The visualization results show that our simi-attention can accurately focuses on the lesions for classification and generate fine segmentation results even for small objects. The overall performances show that our simi-attention can significantly improve the performances of backbone models and outperforms compared attention models on most of datasets for both classification and segmentation.

Contrast- and noise-dependent spatial resolution measurement for deep convolutional neural network-based noise reduction in CT using patient data.

Deep convolutional neural network (DCNN)-based noise reduction methods have been increasingly deployed in clinical CT. Accurate assessment of their spatial resolution properties is required. Spatial resolution is typically measured on physical phantoms, which may not represent the true performance of DCNN in patients as it is typically trained and tested with patient images and the generalizability of DNN to physical phantoms is questionable. In this work, we proposed a patient-data-based framework to measure the spatial resolution of DCNN methods, which involves lesion- and noise-insertion in projection domain, lesion ensemble averaging, and modulation transfer function measurement using an oversampled edge spread function from the cylindrical lesion signal. The impact of varying lesion contrast, dose levels, and CNN denoising strengths were investigated for a ResNet-based DCNN model trained using patient images. The spatial resolution degradation of DCNN reconstructions becomes more severe as the contrast or radiation dose decreased, or DCNN denoising strength increased. The measured 50%/10% MTF spatial frequencies of DCNN with highest denoising strength were (-500 HU:0.36/0.72 mm-1; -100 HU:0.32/0.65 mm-1; -50 HU:0.27/0.53 mm-1; -20 HU:0.18/0.36 mm-1; -10 HU:0.15/0.30 mm-1), while the 50%/10% MTF values of FBP were almost kept constant of 0.38/0.76 mm-1.

Federated Deep Learning With Prototype Matching for Object Extraction From Very-High-Resolution Remote Sensing Images

Deep convolutional neural networks (DCNNs) have become the leading tools for object extraction from very-high-resolution (VHR) remote sensing images. However, the label scarcity problem of local datasets hinders the prediction performances of DCNNs, and privacy concerns regarding remote sensing data often arise in the traditional deep learning schemes. To cope with these problems, we propose a novel federated learning scheme with prototype matching (FedPM) to collaboratively learn a richer DCNN model by leveraging remote sensing data distributed among multiple clients. This scheme conducts federated optimization of DCNNs by aggregating clients’ knowledge in the gradient space without compromising data privacy. Specifically, the prototype matching method is developed to regularize the local training using prototypical representations while reducing the distribution divergence across heterogeneous image data. Furthermore, the derived deviations across local and global prototypes are applied to quantify the effects of local models on the decision boundary and optimize the global model updating via the attention-weighted aggregation scheme. Finally, the sparse ternary compression (STC) method is used to alleviate communication costs. Extensive experimental results derived from VHR aerial and satellite image datasets verify that the FedPM can dramatically improve the prediction performance of DCNNs on object extraction with lower communication costs. To the best of our knowledge, this is the first time federated learning has been applied for remote sensing visual tasks.

Comparing Human and Deep Convolutional Neural Network Performance on Twin Identification

Comparing Human and Deep Convolutional Neural Network Performance on Twin Identification

Recognition of compound characters in Kannada language

&lt;p&gt;&lt;span lang="EN-US"&gt;Recognition of degraded printed compound Kannada characters is a challenging research problem. It has been verified experimentally that noise removal is an essential preprocessing step. Proposed are two methods for degraded Kannada character recognition problem. Method 1 is conventionally used histogram of oriented gradients (HOG) feature extraction for character recognition problem. Extracted features are transformed and reduced using principal component analysis (PCA) and classification performed. Various classifiers are experimented with. Simple compound character classification is satisfactory (more than 98% accuracy) with this method. However, the method does not perform well on other two compound types. Method 2 is deep convolutional neural networks (CNN) model for classification. This outperforms HOG features and classification. The highest classification accuracy is found as 98.8% for simple compound character classification. The performance of deep CNN is far better for other two compound types. Deep CNN turns out to better for pooled character classes.&lt;/span&gt;&lt;/p&gt;

Can deep convolutional neural networks support relational reasoning in the same-different task?

Same-different visual reasoning is a basic skill central to abstract combinatorial thought. This fact has lead neural networks researchers to test same-different classification on deep convolutional neural networks (DCNNs), which has resulted in a controversy regarding whether this skill is within the capacity of these models. However, most tests of same-different classification rely on testing on images that come from the same pixel-level distribution as the training images, yielding the results inconclusive. In this study, we tested relational same-different reasoning in DCNNs. In a series of simulations we show that models based on the ResNet architecture are capable of visual same-different classification, but only when the test images are similar to the training images at the pixel level. In contrast, when there is a shift in the testing distribution that does not change the relation between the objects in the image, the performance of DCNNs decreases substantially. This finding is true even when the DCNNs’ training regime is expanded to include images taken from a wide range of different pixel-level distributions or when the model is trained on the testing distribution but on a different task in a multitask learning context. Furthermore, we show that the relation network, a deep learning architecture specifically designed to tackle visual relational reasoning problems, suffers the same kind of limitations. Overall, the results of this study suggest that learning same-different relations is beyond the scope of current DCNNs.

Learning holistic and discriminative features via an efficient external memory module for building extraction in remote sensing images

In recent years, deep convolutional neural networks (DCNNs) have shown their predominant place in building extraction. Generally, the performance of DCNNs depends on the quality of parameter learning. To improve the performance, many approaches chose to enlarge the network size or employ additional plug-in modules. However, there exists large redundancy in parameter learning of neural networks. We argue that the methods totally relying on parameter learning cannot fully exploit the useful information of the given data, and feature learned by such methods is not comprehensive enough. In this paper, we propose an efficient memory module, named EMM, to enhance the learning ability of DCNNs in building extraction. In our method, we encode each batch of data into memory key-value pairs and store them in memory banks, in which each memory will dynamically update during the training process. Inspired by the non-local module, we propose to use a tri-branch way for a more efficient memory retrieval process. By training with our memory module, DCNNs can learn more holistic and discriminative features efficiently and effectively. We conduct intensive experiments on three challenging public datasets. Compared with popular segmentation networks on all three datasets, our method can achieve competitive performance while keeping relatively fewer parameters.

An improved categorical cross entropy for remote sensing image classification based on noisy labels

In recent years, volunteered geographic information (VGI) is widely used to train deep convolutional neural networks (DCNNs) for high-resolution remote sensing image classification. However, noisy labels are often included in the training samples generated by VGI, which inevitably affects the performance of DCNNs. To solve this problem, the negative effect of noisy labels on remote sensing image classification by DCNNs is analyzed. Then, an improved categorical cross-entropy (ICCE) is proposed to address the issue of noisy labels. The ICCE improves the robustness of DCNN to noisy labels by revisiting the sample weighting scheme so that much attention is paid to the clean samples instead of the noisy samples. Besides, the error bound of ICCE is derived with strict mathematical proof under different types of noisy labels, which ensures the advantages of ICCE from theory. Additionally, extensive experiments are conducted on three remote sensing image datasets with simulated and real noisy labels to quantitatively evaluate the performance of ICCE.

Deep learning prediction of sex on chest radiographs: a potential contributor to biased algorithms.

Deep convolutional neural networks (DCNNs) for diagnosis of disease on chest radiographs (CXR) have been shown to be biased against males or females if the datasets used to train them have unbalanced sex representation. Prior work has suggested that DCNNs can predict sex on CXR, which could aid forensic evaluations, but also be a source of bias. To (1) evaluate the performance of DCNNs for predicting sex across different datasets and architectures and (2) evaluate visual biomarkers used by DCNNs to predict sex on CXRs. Chest radiographs were obtained from the Stanford CheXPert and NIH Chest XRay14 datasets which comprised of 224,316 and 112,120 CXRs, respectively. To control for dataset size and class imbalance, random undersampling was used to reduce each dataset to 97,560 images that were balanced for sex. Each dataset was randomly split into training (70%), validation (10%), and test (20%) sets. Four DCNN architectures pre-trained on ImageNet were used for transfer learning. DCNNs were externally validated using a test set from the opposing dataset. Performance was evaluated using area under the receiver operating characteristic curve (AUC). Class activation mapping (CAM) was used to generate heatmaps visualizing the regions contributing to the DCNN's prediction. On the internal test set, DCNNs achieved AUROCs ranging from 0.98 to 0.99. On external validation, the models reached peak cross-dataset performance of 0.94 for the VGG19-Stanford model and 0.95 for the InceptionV3-NIH model. Heatmaps highlighted similar regions of attention between model architectures and datasets, localizing to the mediastinal and upper rib regions, as well as to the lower chest/diaphragmatic regions. DCNNs trained on two large CXR datasets accurately predicted sex on internal and external test data with similar heatmap localizations across DCNN architectures and datasets. These findings support the notion that DCNNs can leverage imaging biomarkers to predict sex and potentially confound the accurate prediction of disease on CXRs and contribute to biased models. On the other hand, these DCNNs can be beneficial to emergency radiologists for forensic evaluations and identifying patient sex for patients whose identities are unknown, such as in acute trauma.

Microstructure-informed deep convolutional neural network for predicting short-term creep modulus of cement paste

This study aims to provide an efficient alternative for predicting creep modulus of cement paste based on Deep Convolutional Neural Network (DCNN). First, a microscale lattice model for short-term creep is adopted to build a database that contains 18,920 samples. Then, 3 DCNNs with different consecutive convolutional layers are built to learn from the database. Finally, the performance of DCNNs is tested on unseen testing samples. The results show that the DCNNs can achieve high accuracy in the testing set, with the R2 all higher than 0.96. The distribution of creep modulus predicted by the DCNNs coincides with that of the original data. Furthermore, through analyzing the feature maps, it is found that the DCNNs can correctly capture the local importance of different microstructural phases. The DCNN allows therefore prediction of the creep modulus based on microstructural input, which saves computational resources of segmentation procedure and multiple incremental FEM calculations.

Classification of clustered microseismic events in a coal mine using machine learning

Discrimination of seismicity distributed in different areas is essential for reliable seismic risk assessment in mines. Although machine learning has been widely applied in seismic data processing, feasibility and reliability of applying this technique to classify spatially clustered seismic events in underground mines are yet to be investigated. In this research, two groups of seismic events with a minimum local magnitude (ML) of −3 were observed in an underground coal mine. They were respectively located around a dyke and the longwall face. Additionally, two types of undesired signals were also recorded. Four machine learning methods, i.e. random forest (RF), support vector machine (SVM), deep convolutional neural network (DCNN), and residual neural network (ResNN), were used for classifying these signals. The results obtained based on a primary dataset showed that these seismic events could be classified with at least 91% accuracy. The DCNN using seismogram images as the inputs reached the best performance with more than 94% accuracy. As mining is a dynamic progress which could change the characteristics of seismic signals, the temporal variance in the prediction performance of DCNN was also investigated to assess the reliability of this classifier during mining. A cascaded workflow consisting of database update, model training, signal prediction, and results review was established. By progressively calibrating the DCNN model, it achieved up to 99% prediction accuracy. The results demonstrated that machine learning is a reliable tool for the automatic discrimination of spatially clustered seismicity in underground mining.

An Artificial Intelligent Risk Classification Method of High Myopia Based on Fundus Images.

High myopia is a global ocular disease and one of the most common causes of blindness. Fundus images can be obtained in a noninvasive manner and can be used to monitor and follow up on many fundus diseases, such as high myopia. In this paper, we proposed a computer-aided diagnosis algorithm using deep convolutional neural networks (DCNNs) to grade the risk of high myopia. The input images were automatically classified into three categories: normal fundus images were labeled class 0, low-risk high-myopia images were labeled class 1, and high-risk high-myopia images were labeled class 2. We conducted model training on 758 clinical fundus images collected locally, and the average accuracy reached 98.15% according to the results of fivefold cross-validation. An additional 100 fundus images were used to evaluate the performance of DCNNs, with ophthalmologists performing external validation. The experimental results showed that DCNNs outperformed human experts with an area under the curve (AUC) of 0.9968 for the recognition of low-risk high myopia and 0.9964 for the recognition of high-risk high myopia. In this study, we were able to accurately and automatically perform high myopia classification solely using fundus images. This has great practical significance in terms of improving early diagnosis, prevention, and treatment in clinical practice.

Abstract P345: A MRI Displacement-based Deep-learning Semantic Segmentation Tool For Left-ventricular Longitudinal Strain Analysis In Cardiotoxicity

Global longitudinal strain (GLS) computed in the left-ventricle (LV) is an established metric for detecting cardiotoxicity in breast cancer patients treated with antineoplastic agents. The purpose of this study was to develop a novel, MRI-based, deep-learning semantic segmentation tool that automates the phase-unwrapping for LV displacement computation in GLS. Strain analysis via phase-unwrapping was conducted on 30 breast cancer patients and 30 healthy females acquired with the Displacement Encoding with Stimulated Echoes (DENSE) sequence. A ResNet-50 deep convolutional neural network (DCNN) architecture for automated phase-unwrapping, a previously validated ResNet-50 DCNN for chamber quantification and the Radial Point Interpolation Method were used for GLS computation (Figure 1). The DCNN's performance was measured with F1 and Dice scores, and validated in comparison to the robust transport of intensity equation (RTIE) and quality guided phase-unwrapping (QGPU) conventional algorithms. The three techniques were compared by intraclass correlation coefficient with Cronbach’s alpha (C-alpha) index. Classification accuracy with the DCNN was F1 score of 0.92 and Dice score of 0.89. The GLS results from RTIE, QGPU and DCNN were -16.0 ± 2%, -16.1 ± 3% and -15.9 ± 3% (C-alpha = 0.89) for patients and -18.9 ± 3%, -19.0 ± 4% and -18.9 ± 3% (C-alpha = 0.92) for healthy subjects. Comparable validation results from the three techniques show the feasibility of a DCNN-based approach to LV displacement and GLS analysis. The dissimilarities between patients and healthy subjects demonstrate that DCNN-based GLS computation may detect LV abnormalities related to cardiotoxicity.

On the Necessity of Recurrent Processing during Object Recognition: It Depends on the Need for Scene Segmentation.

Although feedforward activity may suffice for recognizing objects in isolation, additional visual operations that aid object recognition might be needed for real-world scenes. One such additional operation is figure-ground segmentation, extracting the relevant features and locations of the target object while ignoring irrelevant features. In this study of 60 human participants (female and male), we show objects on backgrounds of increasing complexity to investigate whether recurrent computations are increasingly important for segmenting objects from more complex backgrounds. Three lines of evidence show that recurrent processing is critical for recognition of objects embedded in complex scenes. First, behavioral results indicated a greater reduction in performance after masking objects presented on more complex backgrounds, with the degree of impairment increasing with increasing background complexity. Second, electroencephalography (EEG) measurements showed clear differences in the evoked response potentials between conditions around time points beyond feedforward activity, and exploratory object decoding analyses based on the EEG signal indicated later decoding onsets for objects embedded in more complex backgrounds. Third, deep convolutional neural network performance confirmed this interpretation. Feedforward and less deep networks showed a higher degree of impairment in recognition for objects in complex backgrounds compared with recurrent and deeper networks. Together, these results support the notion that recurrent computations drive figure-ground segmentation of objects in complex scenes.SIGNIFICANCE STATEMENT The incredible speed of object recognition suggests that it relies purely on a fast feedforward buildup of perceptual activity. However, this view is contradicted by studies showing that disruption of recurrent processing leads to decreased object recognition performance. Here, we resolve this issue by showing that how object recognition is resolved and whether recurrent processing is crucial depends on the context in which it is presented. For objects presented in isolation or in simple environments, feedforward activity could be sufficient for successful object recognition. However, when the environment is more complex, additional processing seems necessary to select the elements that belong to the object and by that segregate them from the background.

Boosting Multilabel Semantic Segmentation for Somata and Vessels in Mouse Brain.

Deep convolutional neural networks (DCNNs) are widely utilized for the semantic segmentation of dense nerve tissues from light and electron microscopy (EM) image data; the goal of this technique is to achieve efficient and accurate three-dimensional reconstruction of the vasculature and neural networks in the brain. The success of these tasks heavily depends on the amount, and especially the quality, of the human-annotated labels fed into DCNNs. However, it is often difficult to acquire the gold standard of human-annotated labels for dense nerve tissues; human annotations inevitably contain discrepancies or even errors, which substantially impact the performance of DCNNs. Thus, a novel boosting framework consisting of a DCNN for multilabel semantic segmentation with a customized Dice-logarithmic loss function, a fusion module combining the annotated labels and the corresponding predictions from the DCNN, and a boosting algorithm to sequentially update the sample weights during network training iterations was proposed to systematically improve the quality of the annotated labels; this framework eventually resulted in improved segmentation task performance. The microoptical sectioning tomography (MOST) dataset was then employed to assess the effectiveness of the proposed framework. The result indicated that the framework, even trained with a dataset including some poor-quality human-annotated labels, achieved state-of-the-art performance in the segmentation of somata and vessels in the mouse brain. Thus, the proposed technique of artificial intelligence could advance neuroscience research.

Application of deep learning to detect Lamb’s quarters (Chenopodium album L.) in potato fields of Atlantic Canada

Excessive use of herbicides for weed control increases the cost of crop production and can lead to environmental degradation. An intelligent spraying system can apply agrochemicals on an as-needed basis by detecting and selectively targeting the weeds. The objective of this research was to investigate the feasibility of using deep convolutional neural networks (DCNNs) for detecting lamb’s quarters (Chenopodium album) in potato fields. Five potato fields were selected in Prince Edward Island (PEI) and New Brunswick (NB), Canada to collect images of spatially and temporally varied potato plants and lamb’s quarters. The image database included pictures, taken under varying growth stages of potato, outdoor light (clear, cloudy, and partly cloudy), and shadowy conditions. The images were trained for DCNN models, namely GoogLeNet, VGG-16, and EfficientNet to classify lamb’s quarters and potato plants. Performance of two frameworks, namely TensorFlow and PyTorch, were compared in training, testing, and during inferring the DCNNs. Results showed excellent performance of DCNNs in lamb’s quarters and potato plant classification (accuracy > 90%). However, the EfficientNet with PyTorch framework showed a maximum accuracy of (0.92–0.97) for every growth stage of the plants. Inference times of DCNNs were recorded using three graphics processing units (GPUs), namely Nvidia GeForce 930MX, Nvidia GeForce GTX1080 Ti, and Nvidia GeForce GTX1050. All the DCNNs performed better with PyTorch than TensorFlow frameworks. It was concluded that the trained models can be used in automation of the spraying systems for the site-specific application of agrochemicals for weed control in potato fields. Such precision agriculture technologies will ensure economically viable and environmentally safe potato cultivation.

Seeing through disguise: Getting to know you with a deep convolutional neural network

People use disguise to look unlike themselves (evasion) or to look like someone else (impersonation). Evasion disguise challenges human ability to see an identity across variable images; Impersonation challenges human ability to tell people apart. Personal familiarity with an individual face helps humans to see through disguise. Here we propose a model of familiarity based on high-level visual learning mechanisms that we tested using a deep convolutional neural network (DCNN) trained for face identification. DCNNs generate a face space in which identities and images co-exist in a unified computational framework, that is categorically structured around identity, rather than retinotopy. This allows for simultaneous manipulation of mechanisms that contrast identities and cluster images. In Experiment 1, we measured the DCNN's baseline accuracy (unfamiliar condition) for identification of faces in no disguise and disguise conditions. Disguise affected DCNN performance in much the same way it affects human performance for unfamiliar faces in disguise (cf. Noyes & Jenkins, 2019). In Experiment 2, we simulated familiarity for individual identities by averaging the DCNN-generated representations from multiple images of each identity. Averaging improved DCNN recognition of faces in evasion disguise, but reduced the ability of the DCNN to differentiate identities of similar appearance. In Experiment 3, we implemented a contrast learning technique to simultaneously teach the DCNN appearance variation and identity contrasts between different individuals. This facilitated identification with both evasion and impersonation disguise. Familiar face recognition requires an ability to group images of the same identity together and separate different identities. The deep network provides a high-level visual representation for face recognition that supports both of these mechanisms of face learning simultaneously.