Inception V3 Convolutional Neural Network Research Articles

Deep Learning Model for Colon Cancer Classification using InceptionV3

Colon cancer is a significant global public health concern, necessitating an accurate and timely diagnosis for effective treatment. Leveraging advancements in deep learning, this study proposes a novel approach to colon cancer classification using InceptionV3 convolutional neural network architecture. A dataset comprising 1600 colonoscopy images divided into colon_aca (adenocarcinoma) and colon_n (normal) classes was utilized. The model demonstrated promising performance, achieving a training accuracy of 98.86% and a validation accuracy of 99.74% after 100 epochs. This success was accomplished by employing a Stochastic Gradient Descent (SGD) optimizer with a learning rate of 0.0001 and momentum of 0.9, along with categorical cross-entropy loss. Our findings underscore the importance of deep learning models, specifically InceptionV3, in facilitating the precise classification of colon cancer, thus offering a valuable tool for assisting clinicians in early detection and treatment decision-making. Future research may explore the integration of additional clinical data and the evaluation of alternative deep learning architectures to further enhance diagnostic accuracy.

Determination of the pubertal growth spurt by artificial intelligence analysis of cervical vertebrae maturation in lateral cephalometric radiographs

ObjectiveThis study aimed to assess the performance of a convolutional neural network (CNN) model in detecting the pubertal growth spurt by analyzing cervical vertebrae maturation (CVM) in lateral cephalometric radiographs (LCRs). Study designIn total, LCRs of patients from 6 to 17 years old were selected. Three radiologists independently and blindly classified the maturation stages on the LCRs and defined the difficulty of each classification. Subsequently, the stage and level of difficulty were determined by consensus. LCRs were distributed between training, validation, and test sets across four CNN-based models. The models’ responses were compared with the radiologists’ reference standard, and the architecture with the highest success rate was selected for evaluation. Models were developed using full and cropped LCRs with original and simplified maturation classifications. ResultsIn the simplified classification, the Inception-v3 CNN yielded accuracy of 74% and 75%, with recall and precision values of 61% and 62%, for full and cropped LCRs, respectively. It achieved 61% and 62% success rates with full and cropped LCRs, respectively, reaching 72.7% for easy-to-classify cropped cases. ConclusionOverall, the CNN model demonstrated potential for determining whether the patient's maturation status regarding the pubertal growth spurt through images of the cervical vertebrae. It may be useful as an initial assessment tool or as an aid for optimizing the assessment and treatment decisions of the clinician.

Utilizing Visual Geometry Group (VGG16) and InceptionV3 convolutional Neural Network (CNN) models for accurate diagnosis of lung cancer: an Artificial Intelligence (AI)-based approach

Utilizing Visual Geometry Group (VGG16) and InceptionV3 convolutional Neural Network (CNN) models for accurate diagnosis of lung cancer: an Artificial Intelligence (AI)-based approach

Prediction of Ovarian Cancer Response to Therapy Based on Deep Learning Analysis of Histopathology Images.

Ovarian cancer remains the leading gynecological cause of cancer mortality. Predicting the sensitivity of ovarian cancer to chemotherapy at the time of pathological diagnosis is a goal of precision medicine research that we have addressed in this study using a novel deep-learning neural network framework to analyze the histopathological images. We have developed a method based on the Inception V3 deep learning algorithm that complements other methods for predicting response to standard platinum-based therapy of the disease. For the study, we used histopathological H&E images (pre-treatment) of high-grade serous carcinoma from The Cancer Genome Atlas (TCGA) Genomic Data Commons portal to train the Inception V3 convolutional neural network system to predict whether cancers had independently been labeled as sensitive or resistant to subsequent platinum-based chemotherapy. The trained model was then tested using data from patients left out of the training process. We used receiver operating characteristic (ROC) and confusion matrix analyses to evaluate model performance and Kaplan-Meier survival analysis to correlate the predicted probability of resistance with patient outcome. Finally, occlusion sensitivity analysis was piloted as a start toward correlating histopathological features with a response. The study dataset consisted of 248 patients with stage 2 to 4 serous ovarian cancer. For a held-out test set of forty patients, the trained deep learning network model distinguished sensitive from resistant cancers with an area under the curve (AUC) of 0.846 ± 0.009 (SE). The probability of resistance calculated from the deep-learning network was also significantly correlated with patient survival and progression-free survival. In confusion matrix analysis, the network classifier achieved an overall predictive accuracy of 85% with a sensitivity of 73% and specificity of 90% for this cohort based on the Youden-J cut-off. Stage, grade, and patient age were not statistically significant for this cohort size. Occlusion sensitivity analysis suggested histopathological features learned by the network that may be associated with sensitivity or resistance to the chemotherapy, but multiple marker studies will be necessary to follow up on those preliminary results. This type of analysis has the potential, if further developed, to improve the prediction of response to therapy of high-grade serous ovarian cancer and perhaps be useful as a factor in deciding between platinum-based and other therapies. More broadly, it may increase our understanding of the histopathological variables that predict response and may be adaptable to other cancer types and imaging modalities.

Hybrid AI models allow label-free identification and classification of pancreatic tumor repopulating cell population

Human pancreatic cancer cell lines harbor a small population of tumor repopulating cells (TRCs). Soft 3D fibrin gel allows efficient selection and growth of these tumorigenic TRCs. However, rapid and high-throughput identification and classification of pancreatic TRCs remain technically challenging. Here, we developed deep learning (DL) models paired with machine learning (ML) models to readily identify and classify 3D fibrin gel-selected TRCs into sub-types. Using four different human pancreatic cell lines, namely, MIA PaCa-2, PANC-1, CFPAC-1, and HPAF-II, we classified 3 main sub-types to be present within the TRC population. Our best model was an Inception-v3 convolutional neural network (CNN) used as a feature extractor paired with a Support Vector Machine (SVM) classifier with radial basis function (rbf) kernel which obtained a test accuracy of 90%. In addition, we compared this hybrid method of supervised classification with other methods of supervised classifications and showed that our working model outperforms others. With the help of unsupervised machine learning algorithms, we also validated that the pancreatic TRC subpopulation can be clustered into 3 sub-types. Collectively, our robust model can detect and readily classify tumorigenic TRC subpopulation label-free in a high-throughput fashion which can be very beneficial in clinical settings.

Developing a convolutional neural network for classifying tumor images using Inception v3

Deep learning algorithms rely on digital pathology to classify tissue tumors, where the whole tissue slides are digitized and imaged. The produced multi-resolution whole slide images (MWSIs) are with high resolution that may range from about 100,000 to 200,000 pixels. MWSIs are often stored in a multi-resolution configuration to simplify the processing of images, navigation, and efficient exposition. This work develops a network for classifying MWSIs that require high memory employing a deep neural Inception-v3 architecture. This work employs the MWSIs from Camelyon16, which is around 451 GB in size of Challenge dataset from two independent sources including 400 MWSIs as a total of lymph nodes. The training dataset contains 111 MWSIs of tumor tissue and lymph nodes and 159 WSIs of normal lymph nodes. The developed model uses sample-based processing to train extensive MWSIs employing the MATLAB platform. The model introduces transfer learning techniques with an Inception-v3-based architecture to categorize separate samples as a tumor or normal. Therefore, the main aim here is to achieve two-classes binary segmentation containing normal and tumor. This includes creating a new fully connected layer for the Inception-v3 architecture with two classes and compensating new layers instead of the original final fully-connected layers. The results obtained demonstrated that the heatmap visualization can recognize the boundary coordinates of ground truth as sketchy Region Of Interest (ROI), where the green boundary represents the normal regions and the tumor area with red boundaries. The proposed Inception v3 Convolutional Neural Network (CNN) architecture can achieve more than 92.8 % accuracy for such MWSIs dataset to categorize brain tumors into normal and tumor tissue

Detecting Driver Drowsiness and Buzzer Alert using CNN Algorithm

At present, driver fatigue has emerged as a predominant root cause of vehicular accidents on the roads. The phenomenon of drowsiness occurring while driving, particularly during twilight hours, can pose a significant risk to road safety, potentially leading to unforeseen collisions resulting in grave injuries or fatalities. The insufficiency of sleep negatively impacts an individual's capacity to carry out tasks efficiently, thereby resulting in prolonged reaction durations, weakened memory, and impaired decision- making abilities. This augmentation accentuates the likelihood of inflicting harm upon fellow riders and pedestrians on the thoroughfare. In response to the aforementioned difficulty, our proposed resolution leverages a Deep Learning methodology employing a convolutional neural network (CNN) framework, with a focus on treating the detection of drowsiness as an object recognition undertaking aimed at discerning and pinpointing the occurrence of opened and closed eyelids. The object detection process in our proposed methodology utilizes the Inception V3 convolutional neural network architecture

Amharic Language Image Captions Generation Using Hybridized Attention-Based Deep Neural Networks

This study aims to develop a hybridized deep learning model for generating semantically meaningful image captions in Amharic Language. Image captioning is a task that combines both computer vision and natural language processing (NLP) domains. However, existing studies in the English language primarily focus on visual features to generate captions, resulting in a gap between visual and textual features and inadequate semantic representation. To address this challenge, this study proposes a hybridized attention-based deep neural network (DNN) model. The model consists of an Inception-v3 convolutional neural network (CNN) encoder to extract image features, a visual attention mechanism to capture significant features, and a bidirectional gated recurrent unit (Bi-GRU) with attention decoder to generate the image captions. The model was trained on the Flickr8k and BNATURE datasets with English captions, which were translated into Amharic Language with the help of Google Translator and Amharic Language experts. The evaluation of the model showed improvement in its performance, with a 1G-BLEU score of 60.6, a 2G-BLEU score of 50.1, a 3G-BLEU score of 43.7, and a 4G-BLEU score of 38.8. Generally, this study highlights the effectiveness of the hybrid approach in generating Amharic Language image captions with better semantic meaning.

Deep time-frequency features and semi-supervised dimension reduction for subject-independent emotion recognition from multi-channel EEG signals

In recent years, human emotion recognition has received great attention since it plays an essential role in human-computer interactions. Traditional methods focused on electroencephalogram(EEG) analysis in either time or frequency domain are unsuitable since EEG signals are nonlinear. This paper proposes subject-independent human emotion recognition from multi-channel EEG signals. The proposed method first obtains the time-frequency content of each channel using the modified Stockwell transform and then extracts the deep features from each time-frequency content by a deep convolutional neural network (CNN). Since there is a huge number of deep features, semi-supervised dimension reduction (SSDR) is utilized to reduce them, and reduced features of all channels are fused to construct the final feature vector. Several CNNs and classifiers are examined respectively for deep feature extraction and classification. The classification considering two-class and four-class scenarios on the DEAP dataset and three-class scenario on the SEED dataset show that the Inception-V3 CNN and support vector machine (SVM) classifier yields the highest accuracy. We present the extensive simulation to present the efficiency of the proposed method. Also, performance comparison with current methods based on time-frequency analysis demonstrates that the proposed method outperforms the others.

PiTLiD: Identification of Plant Disease From Leaf Images Based on Convolutional Neural Network.

With the development of plant phenomics, the identification of plant diseases from leaf images has become an effective and economic approach in plant disease science. Among the methods of plant diseases identification, the convolutional neural network (CNN) is the most popular one for its superior performance. However, CNN's representation power is still a challenge in dealing with small datasets, which greatly affects its popularization. In this work, we propose a new method, namely PiTLiD, based on pretrained Inception-V3 convolutional neural network and transfer learning to identify plant leaf diseases from phenotype data of plant leaf with small sample size. To evaluate the robustness of the proposed method, the experiments on several datasets with small-scale samples were implemented. The results show that PiTLiD performs better than compared methods. This study provides a plant disease identification tool based on a deep learning algorithm for plant phenomics. All the source data and code are accessible at https://github.com/zhanglab-wbgcas/PiTLiD.

Latent Space Representational Learning of Deep Features for Acute Lymphoblastic Leukemia Diagnosis

Acute Lymphoblastic Leukemia (ALL) is a fatal malignancy that is featured by the abnormal increase of immature lymphocytes in blood or bone marrow. Early prognosis of ALL is indispensable for the effectual remediation of this disease. Initial screening of ALL is conducted through manual examination of stained blood smear microscopic images, a process which is time-consuming and prone to errors. Therefore, many deep learning-based computer-aided diagnosis (CAD) systems have been established to automatically diagnose ALL. This paper proposes a novel hybrid deep learning system for ALL diagnosis in blood smear images. The introduced system integrates the proficiency of autoencoder networks in feature representational learning in latent space with the superior feature extraction capability of standard pretrained convolutional neural networks (CNNs) to identify the existence of ALL in blood smears. An augmented set of deep image features are formed from the features extracted by GoogleNet and Inception-v3 CNNs from a hybrid dataset of microscopic blood smear images. A sparse autoencoder network is designed to create an abstract set of significant latent features from the enlarged image feature set. The latent features are used to perform image classification using Support Vector Machine (SVM) classifier. The obtained results show that the latent features improve the classification performance of the proposed ALL diagnosis system over the original image features. Moreover, the classification performance of the system with various sizes of the latent feature set is evaluated. The retrieved results reveal that the introduced ALL diagnosis system superiorly compete the state of the art.

Abstract 1202: A repository of PDX histology images for exploring spatial heterogeneity and cancer dynamics

Abstract Patient-derived xenografts (PDXs) recapitulate intratumoral spatial heterogeneity and simulate a tumor microenvironment in which human immune and stromal cells in the PDX are replaced over passages by murine cells partially lacking immune function. Histological imaging enables exploring the spatial heterogeneity and dynamics of cancer, stromal, and immune cell interactions as correlates of tumor stage and therapeutic response over passages. We created a repository of curated, haematoxylin and eosin (H&E) images as a community resource for addressing these questions. Images were generated at five sites within the NCI’s PDX Development and Trial Centers Research Network (PDXNet) and the NCI Patient-Derived Models Repository. Over 900 images, including 739 from PDXs and 190 from paired patients, are hosted on the Seven Bridges Genomics Cancer Genomics Cloud. They represent 42 cancer subtypes, including breast cancer (n=134), colon adenocarcinoma (COAD; n=94), pancreatic cancer (n=87), lung adenocarcinoma (LUAD; n=80), melanoma (n=71), and squamous cell lung cancer (LUSC; n=65). Paired human/PDX images are available for each of these cancers. Human and/or PDX images generated following patient treatment are available for 37 of the subtypes. Most images are from early passages (P0: 158; P1: 292; P2: 152; P3: 69; >P3: 55). Annotations include sex, age, race, ethnicity, and, for most images, pathological assessment of tissue-level percent cancer, stromal, and necrotic cell content (n=639) and tumor stage (n=650). RNA and exome sequencing data are available for 99 and 228 images, respectively, matched at the patient or sample level. Quality control was performed using HistoQC. Cells were segmented and labeled as neoplastic, necrotic, immune, stromal, or other using Hover-Net and predictions of total neoplastic cell area correlated with whole-slide pathological assessment of cancer cell percentage (COAD: r=0.51; LUSC: r=0.59). HD-Staining, another classification approach, was applied to a subset of images and our clinical annotations will facilitate validation of this and related methods. Features of 512 x 512 pixel tiles were computed using the Inception V3 convolutional neural network pre-trained on ImageNet. Unsupervised clustering of these features demonstrate inter-patient heterogeneity within pathologist-annotated tumor regions. A classifier developed using pathologist-annotated cancer, stromal, and necrotic regions and trained on the features in LUSC images (n=10 images) achieved a cross-validation accuracy of 96% for cancer tiles across (n=5) LUAD images. Accuracy was lower for stromal classification (90%), likely reflecting current limitations of our small, but growing, labeled training set. Our repository of clinically-annotated PDX H&E images should aid the community in studying spatial heterogeneity and in training deep learning-based image analysis methods. Citation Format: Brian S. White, Xingyi Woo, Soner Koc, Todd Sheridan, Steven B. Neuhauser, Akshat M. Savaliya, Lacey E. Dobrolecki, John D. Landua, Matthew H. Bailey, Maihi Fujita, Kurt W. Evans, Bingliang Fang, Junya Fujimoto, Maria Gabriela Raso, Shidan Wang, Guanghua Xiao, Yang Xie, Sherri R. Davies, Ryan C. Fields, R Jay Mashl, Jacqueline L. Mudd, Yeqing Chen, Min Xiao, Xiaowei Xu, Melinda G. Hollingshead, Shahanawaz Jiwani, PDXNet Consortium, Yvonne A. Evrard, Tiffany A. Wallace, Jeffrey A. Moscow, James H. Doroshow, Nicholas Mitsiades, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis G. Carvajal-Carmona, Alana L. Welm, Bryan E. Welm, Michael T. Lewis, Ramaswamy Govindan, Li Ding, Shunqiang Li, Meenhard Herlyn, Michael A. Davies, Jack A. Roth, Funda Meric-Bernstam, Carol J. Bult, Brandi Davis-Dusenbery, Dennis A. Dean, Jeffrey H. Chuang. A repository of PDX histology images for exploring spatial heterogeneity and cancer dynamics [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1202.

0085 SleepInceptionNet: A Deep Learning Algorithm for Real-Time Sleep Stages Scoring Using Single-channel EEG

Abstract Introduction Most of the current automatic polysomnography sleep staging methods use multi-signal data and require a sequence of preceding and following epochs to score the stage of a specific epoch, which may not be desirable for analysis in real-time and/or in free-living conditions. We developed a deep learning-based sleep staging algorithm, namely SleepInceptionNet, that is designed to score each epoch using only single-channel electroencephalogram (EEG) data within that specific epoch. Methods Polysomnography data of 883 participants (937,975 thirty-second epochs) in the Multi-Ethnic Study of Atherosclerosis (MESA) obtained from the National Sleep Research Resource (NSRR) were randomly divided into a separate training/validation set of 194 participants and a test set of 689 participants. Each 30-second raw central EEG channel signal was transformed to time-frequency domain images using continuous wavelet transform method. Sleep stage in each epoch was obtained using SleepInceptionNet, in which the InceptionV3 convolutional neural network structure was trained and tuned on the time-frequency images of the training set to classify each epoch into one of the five stages of Wake, N1, N2, N3, or rapid eye movement (REM) sleep. Results Compared to the ground truth manually scored polysomnography sleep stages, SleepInceptionNet achieved an overall kappa agreement of 0.690 and overall weighted accuracy of 0.897. The model showed accuracy (mean±SD across the test set participants) of 0.940±0.067 in detecting Wake, 0.883±0.047 in detecting N1, 0.845±0.055 in detecting N2, 0.939±0.038 in detecting N3, and 0.930±0.038 in detecting REM epochs, in reference to manually scored polysomnography. Conclusion SleepInceptionNet showed a high agreement with manually scored polysomnography in epoch-by-epoch classification of sleep stages. This study demonstrates the viability of real-time, accurate sleep staging using a single-channel EEG, which could have a variety of applications such as delivery of on-demand interventions during specific sleep stages in free-living conditions. Support (If Any) MESA Sleep Ancillary study was funded by NIH-NHLBI Association of Sleep Disorders with Cardiovascular Health Across Ethnic Groups (RO1 HL098433). MESA is supported by NHLBI (HHSN268201500003I, N01-HC-95159, N01-HC-95160, N01-HC-95161, N01-HC-95162, N01-HC-95163, N01-HC-95164, N01- HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168, and N01-HC-95169) and NCATS (UL1-TR-000040, UL1-TR-001079, and UL1-TR-001420). NSRR was supported by NHLBI (R24 HL114473, 75N92019R002).

YOLO Object Detector and Inception-V3 Convolutional Neural Network for Improved Brain Tumor Segmentation

For a range of medical analysis applications, the localization of brain tumors and brain tumor segmentation from magnetic resonance imaging (MRI) are challenging yet critical jobs. Many recent studies have included four modalities: i.e., T1, T1c, T2 & FLAIR, it is because every tumor causing area can be detailed examined by each of these brain imaging modalities. Although the BRATS 2018 datasets give impressive segmentation results, the results are still more complex and need more testing and more training. That’s why this paper recommends operated pre-processing strategies on a small part of an image except for a full image because that’s how an effective and flexible segmented system of brain tumor can be created. In the first phase, an ensemble classification model is developed using different classifiers such as decision tree, SVM, KNN etc. to classify an image into the tumor and non-tumor class by using the strategy of using a small section can completely solve the over-fitting problems and reduces the processing time in a model of YOLO object detector using inceptionv3 CNN features. The second stage is to recommend an efficient and basic Cascade CNN (C-ConvNet/C-CNN), as we deal with a tiny segment of the brain image in each and every slice. In two independent ways, the Cascade-Convolutional Neutral Network model extracts learnable features. On the dataset of BRATS 2018, BRATS 2019 and BRATS 2020, the extensive experimental task has been carried out on the proposed tumor localization framework.: the IoU score achieved of three datasets are 97%, 98% and 100%. Other qualitative evaluations & quantitative evaluations are discussed and presented in the manuscript in detail.

A tandem segmentation-classification approach for the localization of morphological predictors of C. elegans lifespan and motility.

C. elegans is an established model organism for studying genetic and drug effects on aging, many of which are conserved in humans. It is also an important model for basic research, and C. elegans pathologies is a new emerging field. Here we develop a proof-of-principal convolutional neural network-based platform to segment C. elegans and extract features that might be useful for lifespan prediction. We use a dataset of 734 worms tracked throughout their lifespan and classify worms into long-lived and short-lived. We designed WormNet - a convolutional neural network (CNN) to predict the worm lifespan class based on young adult images (day 1 – day 3 old adults) and showed that WormNet, as well as, InceptionV3 CNN can successfully classify lifespan. Based on U-Net architecture we develop HydraNet CNNs which allow segmenting worms accurately into anterior, mid-body and posterior parts. We combine HydraNet segmentation, WormNet prediction and the class activation map approach to determine the segments most important for lifespan classification. Such a tandem segmentation-classification approach shows the posterior part of the worm might be more important for classifying long-lived worms. Our approach can be useful for the acceleration of anti-aging drug discovery and for studying C. elegans pathologies.

A deep learning approach based on stochastic gradient descent and least absolute shrinkage and selection operator for identifying diabetic retinopathy

More than eighty-five to ninety percentage of the diabetic patients are affected with diabetic retinopathy (DR) which is an eye disorder that leads to blindness. The computational techniques can support to detect the DR by using the retinal images. However, it is hard to measure the DR with the raw retinal image. This paper proposes an effective method for identification of DR from the retinal images. In this research work, initially the Weiner filter is used for preprocessing the raw retinal image. Then the preprocessed image is segmented using fuzzy c-mean technique. Then from the segmented image, the features are extracted using grey level co-occurrence matrix (GLCM). After extracting the fundus image, the feature selection is performed stochastic gradient descent, and least absolute shrinkage and selection operator (LASSO) for accurate identification during the classification process. Then the inception v3-convolutional neural network (IV3-CNN) model is used in the classification process to classify the image as DR image or non-DR image. By applying the proposed method, the classification performance of IV3-CNN model in identifying DR is studied. Using the proposed method, the DR is identified with the accuracy of about 95%, and the processed retinal image is identified as mild DR.

A Benchmark for Feature-injection Architectures in Image Captioning

Describing an image with a grammatically and semantically correct sentence, known as image captioning, has been improved significantly with recent advances in computer vision (CV) and natural language processing (NLP) communities. The integration of these communities leads to the development of feature-injection architectures, which define how extracted features are used in captioning. In this paper, a benchmark of feature-injection architectures that utilize CV and NLP techniques is reported for encoder-decoder based captioning. Benchmark evaluations include Inception-v3 convolutional neural network to extract image features in the encoder while the feature-injection architectures such as init-inject, pre-inject, par-inject and merge are applied with a multi-layer gated recurrent unit (GRU) to generate captions in the decoder. Architectures have been evaluated extensively on the MSCOCO dataset across eight performance metrics. It has been concluded that the init-inject architecture with 3-layer GRU outperforms the other architectures in terms of captioning accuracy.

Automated Individual Cattle Identification Using Video Data: A Unified Deep Learning Architecture Approach

Individual cattle identification is a prerequisite and foundation for precision livestock farming. Existing methods for cattle identification require radio frequency or visual ear tags, all of which are prone to loss or damage. Here, we propose and implement a new unified deep learning approach to cattle identification using video analysis. The proposed deep learning framework is composed of a Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) with a self-attention mechanism. More specifically, the Inception-V3 CNN was used to extract features from a cattle video dataset taken in a feedlot with rear-view. Extracted features were then fed to a BiLSTM layer to capture spatio-temporal information. Then, self-attention was employed to provide a different focus on the features captured by BiLSTM for the final step of cattle identification. We used a total of 363 rear-view videos from 50 cattle at three different times with an interval of 1 month between data collection periods. The proposed method achieved 93.3% identification accuracy using a 30-frame video length, which outperformed current state-of-the-art methods (Inception-V3, MLP, SimpleRNN, LSTM, and BiLSTM). Furthermore, two different attention schemes, namely, additive and multiplicative attention mechanisms were compared. Our results show that the additive attention mechanism achieved 93.3% accuracy and 91.0% recall, greater than multiplicative attention mechanism with 90.7% accuracy and 87.0% recall. Video length also impacted accuracy, with video sequence length up to 30-frames enhancing identification performance. Overall, our approach can capture key spatio-temporal features to improve cattle identification accuracy, enabling automated cattle identification for precision livestock farming.

Social media influencers, product placement and network engagement: using AI image analysis to empirically test relationships

PurposeThis research tests empirically the level of consumer engagement with a product via a nonbrand-controlled platform. The authors explore how social media influencers and traditional celebrities are using products within their own social media Instagram posts and how well their perceived endorsement of that product engages their network of followers.Design/methodology/approachA total of 226,881 posts on Instagram were analyzed using the Inception V3 convolutional neural network (CNN) pre-trained on the ImageNet dataset to identify product placement within the Instagram images of 75 of the world's most important social media influencers. The data were used to empirically test the relationships between influencers, product placement and network engagement and efficiency.FindingsInfluencers achieved higher network engagement efficiencies than celebrities; however, celebrity reach was important for engagement overall. Specialty influencers, known for their “subject” expertise, achieved better network engagement efficiency for related product categories. The highest level of engagement efficiency was achieved by beauty influencers advocating and promoting cosmetic and beauty products.Practical implicationsTo maximize engagement and return on investment, manufacturers, retailers and brands must ensure a close fit between the product type and category of influencer promoting a product within their social media posts.Originality/valueMost research to date has focused on brand-controlled social media accounts. This study focused on traditional celebrities and social media influencers and product placement within their own Instagram posts to extend understanding of the perception of endorsement and subsequent engagement with followers. The authors extend the theory of network effects to reflect the complexity inherent in the context of social media influencers.

Deep Gated Recurrent Unit for Smartphone-Based Image Captioning

Expressing the visual content of an image in natural language form has gained relevance due to technological and algorithmic advances together with improved computational processing capacity. Many smartphone applications for image captioning have been developed recently as built-in cameras provide advantages of easy-operation and portability, resulting in capturing an image whenever or wherever needed. Here, an encoder-decoder framework based new image captioning approach with a multi-layer gated recurrent unit is proposed. The Inception-v3 convolutional neural network is employed in the encoder due to its capability of more feature extraction from small regions. The proposed recurrent neural network-based decoder utilizes these features in the multi-layer gated recurrent unit to produce a natural language expression word-by-word. Experimental evaluations on the MSCOCO dataset demonstrate that our proposed approach has the advantage over existing approaches consistently across different evaluation metrics. With the integration of the proposed approach to our custom-designed Android application, named “VirtualEye+”, it has great potential to implement image captioning in daily routine.