Computer Vision Models Research Articles

Automatic Switching of Electric Locomotive Power in Railway Neutral Sections Using Image Processing.

This article presents a computer vision-based approach to switching electric locomotive power supplies as the vehicle approaches a railway neutral section. Neutral sections are defined as a phase break in which the objective is to separate two single-phase traction supplies on an overhead railway supply line. This separation prevents flashovers due to high voltages caused by the locomotives shorting both electrical phases. The typical system of switching traction supplies automatically employs the use of electro-mechanical relays and induction magnets. In this paper, an image classification approach is proposed to replace the conventional electro-mechanical system with two unique visual markers that represent the 'Open' and 'Close' signals to initiate the transition. When the computer vision model detects either marker, the vacuum circuit breakers inside the electrical locomotive will be triggered to their respective positions depending on the identified image. A Histogram of Oriented Gradient technique was implemented for feature extraction during the training phase and a Linear Support Vector Machine algorithm was trained for the target image classification. For the task of image segmentation, the Circular Hough Transform shape detection algorithm was employed to locate the markers in the captured images and provided cartesian plane coordinates for segmenting the Object of Interest. A signal marker classification accuracy of 94% with 75 objects per second was achieved using a Linear Support Vector Machine during the experimental testing phase.

Abdominal MRI Unconditional Synthesis with Medical Assessment

Current computer vision models require a significant amount of annotated data to improve their performance in a particular task. However, obtaining the required annotated data is challenging, especially in medicine. Hence, data augmentation techniques play a crucial role. In recent years, generative models have been used to create artificial medical images, which have shown promising results. This study aimed to use a state-of-the-art generative model, StyleGAN3, to generate realistic synthetic abdominal magnetic resonance images. These images will be evaluated using quantitative metrics and qualitative assessments by medical professionals. For this purpose, an abdominal MRI dataset acquired at Garcia da Horta Hospital in Almada, Portugal, was used. A subset containing only axial gadolinium-enhanced slices was used to train the model. The obtained Fréchet inception distance value (12.89) aligned with the state of the art, and a medical expert confirmed the significant realism and quality of the images. However, specific issues were identified in the generated images, such as texture variations, visual artefacts and anatomical inconsistencies. Despite these, this work demonstrated that StyleGAN3 is a viable solution to synthesise realistic medical imaging data, particularly in abdominal imaging.

Do deep learning models accurately measure visual destination image? A comparison of a fine-tuned model to past work

The measurement of destination image from visual media such as online photography is of growing significance to destination managers and marketers who want to make better decisions and attract more visitors to their destination. However, there is no single approach with proven accuracy for doing this. We present a new approach where we fine-tune a deep learning model for a predetermined set of cognitive attributes of destination image. We then train state of the art neural networks using labelled tourist photography and test accuracy by comparing results with a ground truth dataset built for the same set of visual classes. Comparing our fine-tuned model against results which follow past approaches, we demonstrate that the pre-trained models without fine-tuning are not as accurate in capturing all of the destination image’s cognitive attributes. This is, to the best of our knowledge, the first deep learning computer vision model trained specifically to measure the cognitive component of destination image from photography and can act as a benchmark for future systems.

The Communal Roots of Mexico’s Maquila Industry: Urbanization, Land, and Inequality in Ciudad Juárez, 1960–2000

Abstract This article analyzes the urbanization and privatization of communal lands (ejidos) in Ciudad Juárez, Mexico, via a computer vision model that utilizes Google Street View (GSV) and Geographic Information System (GIS) imagery. Our innovative methodology reveals how processes of ejidal urbanization in Mexico’s northern borderlands contributed to the rise of multinational factories (maquiladoras) and geographies of inequality and violence. Past scholarship on the (d)evolution of ejidal land tenure details how periurban ejidal lands throughout Mexico were often sites of impoverishment and a lack of investment, featuring informal settlement and rapid or chaotic urbanization following the country’s 1960 urban turn. Through its use of novel sources and methods, this article demonstrates that the urbanization process in Juárez’s principal periurban ejidos diverged from this classic model in specific ways. By combining conventional historical sources with visual data like GSV and GIS imagery, Juárez’s former and current ejidal landscapes reveal high levels of investment, formal planning, and infrastructure. We argue that Juárez’s distinctive physical, political, and economic geographies shaped the overwhelmingly industrial, private, and invested character of the city’s (former and current) periurban ejidal lands. This process occurred via a globalized modernization regime that forged disparate landscapes of investment and inequality beginning in the 1950s.

Development of a video-based deep learning model for differentiation of malignant and benign lesions during staging laparoscopy: Is the machine better than the expert?

e13616 Background: Laparoscopic staging of abdominal cancer is routinely performed to assess the presence of peritoneal metastasis (PM). One major challenge of laparoscopic staging is the ability to distinguish between a malignant and benign peritoneal lesion, particularly scar tissue. Lesions' recognition is subjective to intra-case variability, surgeons' experience, type of primary tumor and response to systemic chemotherapy. The Peritoneal Regression Grading Score (PRGS) offers an objective histologic evaluation of the biopsied lesion by assessing the complete or partial presence of malignant cells (PRGS 2-3-4) and fibrous or benign lesion (PRGS 1). Machine learning-based (ML-based) computer vision has various applications in the medical field and can be useful in deducing clinical information through surgical video analysis. The aim of this study was to develop a ML model that can aid the surgeon in the intra-operative assessment of PM. Methods: Retrospectively collected videos of laparoscopies for the staging of PM before cytoreductive surgery or PIPAC were screened from the institutional database for the presence of recorded biopsies. PRGS from the pathology report of these biopsies was retrieved and reviewed by a pathologist. The surgical phase of the biopsy was annotated by systematic selection of 200 frames before the closure of the biopsy instrument on the tissue. One single frame was selected from each timeframe based on its representativeness. Two trained annotators performed the selection, one surgical expert reviewed each annotation. Discrepancies were solved by consensus. ML models based on multiple (ResEfficientnetV2_L) and single frames (Resnet50) per biopsy were trained on the PRGS using a one-versus-all binary classification. Five-fold cross-validation was applied. Two oncologic surgeons, experts in treating and assessing PM, blindly and independently scored the biopsies with PRGS. Results: A total of 127 videos from 67 patients were identified, showcasing PM of gastric, colorectal, appendix, hepatic, gallbladder, breast carcinomas, primary peritoneal tumors and benign lesions. Annotation was performed on 463 biopsies: PRGS 1: 204, PRGS 2: 164, PRGS 3: 70, PRGS 4: 25. Based on timeframe annotation 5214 (PRGS 1: 2044, PRGS 2: 1814, PRGS 3: 929, PRGS 4: 427) images were identified. The model trained on one-versus-all classification for single and multiple images showed an accuracy of 52.4% (Precision 56.2, Recall 48.2) and 45.7% (Precision 76.8, Recall 36.0) respectively in the validation set. Accuracy of each expert was 36.0% and 30.5%. Conclusions: While the accuracy of computer vision models may not be clinically acceptable as a substitute for biopsies, they outperform surgeons in differentiating PM from benign lesions. Computer vision models might be of aid for the surgeon in the recognition of these lesions.

Comparative study of teachable machine for forest fire and smoke detection by drone

Forests play a vital role in maintaining ecological equilibrium and serving as vital habitats for wildlife. They regulate global climate, safeguard soil and water resources, and provide crucial ecosystem services such as air and water purification, essential for human well-being and sustainable development. Forest fires wreak havoc on ecosystems and wildlife, emitting harmful pollutants, disrupting communities, and increasing the risk of erosion and landslides. Detecting forest fires through satellite imaging, aerial reconnaissance, and ground-based sensors is pivotal for early detection and containment, safeguarding human lives, wildlife, and preserving natural resources for future generations. Utilizing drones and deep learning (DL) algorithms can significantly enhance early fire detection and minimize their devastating impact. In this paper, we examine teachable machine, a Google tool for creating DL models. We compare the top model generated by teachable machine for fire and smoke detection to models obtained through transfer learning from established DL models in image recognition and computer vision (CV), such as VGG16, VGG19, MobileNet, MobileNetv2, and MobileNetv3. The results underscore the significance of employing the teachable machine model in specific fire and smoke detection scenarios.

CONVOLUTIONAL NEURAL NETWORKS USAGE FOR RASTER IMAGE PROCESSING IN NON-DESTRUCTIVE TESTING

An overview of Russian and foreign literature sources on the use of computer vision technologies and convolutional artificial neural networks in technological operations of visual control is presented. Methods and algorithms review of computer vision and artificial neural network models used by the authors of scientific articles, methods of preprocessing and expanding the initial data set for models training under consideration and verifying the correctness of machine vision algorithms is carried out. Possible approaches for automating the assessment of visual quality control in various fields of industry are highlighted, research results, results of machine vision algorithms, neural network models, their accuracy and efficiency indicators are presented, as well as areas and possible scenarios for the application of the presented algorithmic solutions. Some results of machine vision algorithms, neural network models, their accuracy and efficiency indicators are presented.

Simultaneous Classification and Location of Volcanic Deformation in SAR Interferograms Using a Convolutional Neural Network

AbstractWith the evolution of interferometric synthetic aperture radar into a tool for active hazard monitoring, new methods are sought to quickly and automatically interpret the large number of interferograms that are created. We present a convolutional neural network (CNN) that is able to both classify the type of deformation, and to locate the deformation within an interferogram in a single step. We achieve this through building a “two headed model,” which returns both outputs after one forward pass of an interferogram through the network. We train our model by first creating a data set of synthetic interferograms, but find that our model's performance is improved through the inclusion of real Sentinel‐1 data. We also investigate how model performance can be improved by best organizing interferograms such that they can exploit the three channel nature of computer vision models trained on very large databases of labeled color images, but find that using different data in each of the three input channels degrades performance when compared to the simple case of repeating wrapped or unwrapped phase across each channel. We also release our labeled Sentinel‐1 interferograms as a database named VolcNet, which consists of ∼500,000 labeled interferograms. VolcNet comprises of time series of unwrapped phase and labels of the magnitude, location, and duration of deformation, which allows for the automatic creation of interferograms between any two acquisitions, and greatly increases the amount of data available compared to other labeling strategies.

Real-Time Fire Surveillance with Machine Learning Twilio Integration

Fire threats pose significant risks to people and property, necessitating efficient surveillance systems. A complete fire surveillance system is proposed using machine learning and artificial intelligence to monitor fire incidents in real time. The system is built using React.js for frontend development and MongoDB for backend storage. Node.js is integrated for server-side operations, ensuring data management and user interaction. The system sends alerts via WhatsApp when image analysis identifies a fire concern, leveraging Twilio for seamless messaging. Robo Flow simplifies computer vision model management, while YOLOv8, a cutting-edge object recognition algorithm, enhances detection speed and accuracy. YOLOv8 is widely used in real-time object identification applications like robotics, autonomous cars and surveillance systems. Twilio is a cloud communication platform that allows developers to integrate voice, video, and SMS into their apps, enabling notifications, alarms, and two-way communication. The research utilizes Machine Learning, React.js, MongoDB, YOLOv8, and Twilio to offer efficient real-time fire surveillance.

Deep learning for identifying bee species from images of wings and pinned specimens.

One of the most challenging aspects of bee ecology and conservation is species-level identification, which is costly, time consuming, and requires taxonomic expertise. Recent advances in the application of deep learning and computer vision have shown promise for identifying large bumble bee (Bombus) species. However, most bees, such as sweat bees in the genus Lasioglossum, are much smaller and can be difficult, even for trained taxonomists, to identify. For this reason, the great majority of bees are poorly represented in the crowdsourced image datasets often used to train computer vision models. But even larger bees, such as bumble bees from the B. vagans complex, can be difficult to separate morphologically. Using images of specimens from our research collections, we assessed how deep learning classification models perform on these more challenging taxa, qualitatively comparing models trained on images of whole pinned specimens or on images of bee forewings. The pinned specimen and wing image datasets represent 20 and 18 species from 6 and 4 genera, respectively, and were used to train the EfficientNetV2L convolutional neural network. Mean test precision was 94.9% and 98.1% for pinned and wing images respectively. Results show that computer vision holds great promise for classifying smaller, more difficult to identify bees that are poorly represented in crowdsourced datasets. Images from research and museum collections will be valuable for expanding classification models to include additional species, which will be essential for large scale conservation monitoring efforts.

Comparative analysis of computer vision algorithms for the real-time detection of digital dermatitis in dairy cows

Digital dermatitis (DD) is a bovine claw disease responsible for ulcerative lesions on the planar aspect of the hoof. DD is associated with massive herd outbreaks of lameness and influences cattle welfare and production. Early detection of DD can lead to prompt treatment and decrease lameness. Computer vision (CV) provides a unique opportunity to improve early detection. The study aims to train and compare applications for the real-time detection of DD in dairy cows. Eight CV models were trained for detection and scoring, compared using performance metrics and inference time, and the best model was automated for real-time detection using images and video. Images were collected from commercial dairy farms while facing the interdigital space on the plantar surface of the foot. Images were scored for M-stages of DD by a trained investigator using the M-stage DD classification system with distinct labels for hyperkeratosis (H) and proliferations (P). Two sets of images were compiled: the first dataset (Dataset 1) containing 1,177 M0/M4H and 1,050 M2/M2P images and the second dataset (Dataset 2) containing 240 M0, 17 M2, 51 M2P, 114 M4H, and 108 M4P images. Models were trained to detect and score DD lesions and compared for precision, recall, and mean average precision (mAP) in addition to inference time in frame per second (FPS). Seven of the nine CV models performed well compared to the ground truth of labeled images using Dataset 1. The six models, Faster R-CNN, Cascade R-CNN, YOLOv3, Tiny YOLOv3, YOLOv4, Tiny YOLOv4, and YOLOv5s achieved an mAP between 0.964 and 0.998, whereas the other two models, SSD and SSD Lite, yielded an mAP of 0.371 and 0.387 respectively. Overall, YOLOv4, Tiny YOLOv4, and YOLOv5s outperformed all other models with almost perfect precision, perfect recall, and a higher mAP. Tiny YOLOv4 outperformed all other models with respect to inference time at 333 FPS, followed by YOLOv5s at 133 FPS and YOLOv4 at 65 FPS. YOLOv4 and Tiny YOLOv4 performed better than YOLOv5s compared to the ground truth using Dataset 2. YOLOv4 and Tiny YOLOv4 yielded a similar mAP of 0.896 and 0.895, respectively. However, Tiny YOLOv4 achieved both higher precision and recall compared to YOLOv4. Finally, Tiny YOLOv4 was able to detect DD lesions on a commercial dairy farm with high performance and speed. The proposed CV tool can be used for early detection and prompt treatment of DD in dairy cows. This result is a step towards applying CV algorithms to veterinary medicine and implementing real-time DD detection on dairy farms.

Variable Temporal Length Training for Action Recognition CNNs.

Most current deep learning models are suboptimal in terms of the flexibility of their input shape. Usually, computer vision models only work on one fixed shape used during training, otherwise their performance degrades significantly. For video-related tasks, the length of each video (i.e., number of video frames) can vary widely; therefore, sampling of video frames is employed to ensure that every video has the same temporal length. This training method brings about drawbacks in both the training and testing phases. For instance, a universal temporal length can damage the features in longer videos, preventing the model from flexibly adapting to variable lengths for the purposes of on-demand inference. To address this, we propose a simple yet effective training paradigm for 3D convolutional neural networks (3D-CNN) which enables them to process videos with inputs having variable temporal length, i.e., variable length training (VLT). Compared with the standard video training paradigm, our method introduces three extra operations during training: sampling twice, temporal packing, and subvideo-independent 3D convolution. These operations are efficient and can be integrated into any 3D-CNN. In addition, we introduce a consistency loss to regularize the representation space. After training, the model can successfully process video with varying temporal length without any modification in the inference phase. Our experiments on various popular action recognition datasets demonstrate the superior performance of the proposed method compared to conventional training paradigm and other state-of-the-art training paradigms.

Understanding climate-related visual storytelling on TikTok

This cross-cultural study investigates the prevalence and impact of climate-related campaigns on TikTok, with a specific focus on climate-related visual storytelling in Indonesia, Japan, Pakistan, the Philippines, Thailand, the United Kingdom, and the United States. Computational methods are employed in the study to analyse a dataset of 7,564 videos, providing insights into prominent visual characteristics and regional variations. The findings underline the significance of cultural and political contexts in shaping climate storytelling on TikTok. Furthermore, this research explores the potential of computational visual data analysis in studying climate communication, demonstrating the integration of computer vision and topic modelling to examine visual styles and communicative functions in TikTok’s climate storytelling. The study enhances our understanding of climate communication on digital platforms and emphasizes the value of leveraging computational methods to gain meaningful cross-cultural insights into visual storytelling in the context of climate change.

Robustness of Deep Learning models in electrocardiogram noise detection and classification

Background and Objective: Automatic electrocardiogram (ECG) signal analysis for heart disease detection has gained significant attention due to busy lifestyles. However, ECG signals are susceptible to noise, which adversely affects the performance of ECG signal analysers. Traditional blind filtering methods use predefined noise frequency and filter order, but they alter ECG biomarkers. Several Deep Learning-based ECG noise detection and classification methods exist, but no study compares recurrent neural network (RNN) and convolutional neural network (CNN) architectures and their complexity. Methods:This paper introduces a knowledge-based ECG filtering system using Deep Learning to classify ECG noise types and compare popular computer vision model architectures in a practical Internet of Medical Things (IoMT) framework. Experimental results demonstrate that the CNN-based ECG noise classifier outperforms the RNN-based model in terms of performance and training time. Results:The study shows that AlexNet, visual geometry group (VGG), and residual network (ResNet) achieved over 70% accuracy, specificity, sensitivity, and F1 score across six datasets. VGG and ResNet performances were comparable, but VGG was more complex than ResNet, with only a 4.57% less F1 score. Conclusions:This paper introduces a Deep Learning (DL) based ECG noise classifier for a knowledge-driven ECG filtering system, offering selective filtering to reduce signal distortion. Evaluation of various CNN and RNN-based models reveals VGG and Resnet outperform. Further, the VGG model is superior in terms of performance. But Resnet performs comparably to VGG with less model complexity.

Intelligent Infrastructure for Traffic Monitoring Based on Deep Learning and Edge Computing

In the field of traffic management and control systems, we are witnessing a symbiotic evolution, where intelligent infrastructure is progressively collaborating with smart vehicles to produce benefits for traffic monitoring and security, by rapidly identifying hazardous behaviours. This exponential growth is due to the rapid development of deep learning in recent years, as well as the improvements in computer vision models. These technologies allow for monitoring tasks without the need to install numerous sensors or stop the traffic, using the extensive camera network of surveillance cameras already present in worldwide roads. This study proposes a computer vision-based solution that allows for real-time processing of video streams through edge computing devices, eliminating the need for Internet connectivity or dedicated sensors. The proposed system employs deep learning algorithms and vision techniques that perform vehicle detection, classification, tracking, speed estimation, and vehicle geolocation.

Big data analytics for image processing and computer vision technologies in sports health management.

Visualization of sports has a lot of potential for future development in data sports because of how quickly things are changing and how much sports depend on data. Presently, conventional systems fail to accurately address sports persons' dynamic health data change with less error rate. Further, those systems are unable to distinguish players' health data and their visualization in a precise manner. An excellent starting point for building fitness solutions based on computer vision technology is the data visualization technology that arose in the age of big data analytics. This research presents a Big Data Analytic assisted Computer Vision Model (BD-CVM) for effective sports persons healthcare data management with improved accuracy and precision. The fitness and health of professional athletes are analyzed using information from a publiclyavailable sports visualization dataset. Machine learning-assisted computer vision dynamic algorithm has been used for an effective image featuring and classification by categorizing sports videos through temporal and geographical data. The significance of big data's great potential in screening data during a sporting event can be reasonably analyzed and processed effectively with less error rate. The proposed BD-CVM utilized an error analysis module which can be embedded in the design further to ensure the accuracy requirements in the data processing from sports videos. The research findings of this paper demonstrate that the strategy presented here can potentially improve accuracy and precision and optimize mean square error in sports data classification and visualization.

Use of Artificial Intelligence to Monitor the Reliability of Removable Load-Handling Devices

Introduction. The malfunction of removable load-handling devices (RLHD) poses significant production risks. That is why research in this field is relevant. The problem has often become a topic of scientific investigation. The authors propose using artificial intelligence more extensively to monitor the state of RLHD. This paper presents a study on how to improve the machine vision model to better identify the absence of locks on RLHD hooks. A probable occurrence of such an issue in production is noted. A storage and monitoring system for RLHD condition is proposed. The aim of this study is to demonstrate the potential for further training of neural networks to significantly enhance the efficiency of RLHD monitoring, ensuring their safe use.Materials and Methods. The work is based on the results of a survey conducted at the LLC “KZ Rostselmash” plant from 2022 to 2023, involving 144 RLHD. Mathematical statistics methods were used to process the data. A neural network model previously trained using the YOLO computer vision algorithm was studied. It was retrained taking into account the norms of the rejection of RLHD, specified in federal rules and standards. Images of RLHD with defects and missing parts were collected from these sources and used to create a training database. The database was expanded by augmentation. The Roboflow platform was used for work.Results. The array of images used for further training of the neural network was divided into three samples: training (88%), validation (8%) and test (4%). These samples were used to train and validate its results. The training was completed after 260 epochs, with a steady increase in accuracy. The neural network model of computer vision obtained in this way automatically detected a common defect in the RLHD hook — the absence of a lock. Its performance was assessed using three indicators: average accuracy (94%), prediction accuracy (88.8%) and response (89.2%). The neural network could receive images from a video camera in real-time and recognize hook defects. During the RLHD inspection at the Rostselmash plant, a grab for lifting engines was found to have all three hooks defective — without locks. To avoid such situations, at the end of work, it was recommended to place the RLHD on a special stand equipped with a microcontroller device that could monitor for a number of potential issues using radio frequency identification.Discussion and Conclusions. The main goal of this proposed solution is to detect and address signs of non-compliance with the established standards. This task can be implemented in facilities that use lifting equipment. In this case, the timely noticed RLHD defects will allow preventing production incidents. As a result, material damage can be reduced and injury statistics improved.

Synthetic Data for Video Surveillance Applications of Computer Vision: A Review

In recent years, there has been a growing interest in synthetic data for several computer vision applications, such as automotive, detection and tracking, surveillance, medical image analysis and robotics. Early use of synthetic data was aimed at performing controlled experiments under the analysis by synthesis approach. Currently, synthetic data are mainly used for training computer vision models, especially deep learning ones, to address well-known issues of real data, such as manual annotation effort, data imbalance and bias, and privacy-related restrictions. In this work, we survey the use of synthetic training data focusing on applications related to video surveillance, whose relevance has rapidly increased in the past few years due to their connection to security: crowd counting, object and pedestrian detection and tracking, behaviour analysis, person re-identification and face recognition. Synthetic training data are even more interesting in this kind of application, to address further, specific issues arising, e.g., from typically unconstrained image or video acquisition conditions and cross-scene application scenarios. We categorise and discuss the existing methods for creating synthetic data, analyse the synthetic data sets proposed in the literature for each of the considered applications, and provide an overview of their effectiveness as training data. We finally discuss whether and to what extent the existing synthetic data sets mitigate the issues of real data, highlight existing open issues, and suggest future research directions in this field.

Deploying a Computer Vision Model Based on YOLOv8 Suitable for Drones in the Tuna Fishing and Aquaculture Industry

In recent years, the global tuna fishing and aquaculture industry has encountered significant challenges in balancing operational efficiency with sustainable resource management. This study introduces an innovative approach utilizing an advanced computer vision model, PA-YOLOv8, specifically adapted for drones, to enhance the monitoring and management of tuna populations. PA-YOLOv8 leverages the capabilities of YOLOv8, a state-of-the-art object detection system known for its precision and speed, tailored to address the unique demands of aerial surveillance in marine environments. Through comprehensive modifications including downsampling techniques, feature fusion enhancements, and the integration of the Global Attention Module (GAM), the model significantly improves the detection accuracy of small and juvenile tuna within complex aquatic landscapes. Experimental results using the Tuna dataset from Roboflow demonstrate marked improvements in detection metrics such as precision, recall, and mean average precision (mAP), affirming the model’s effectiveness. This study underscores the potential of integrating cutting-edge technologies like UAVs and computer vision in promoting sustainable practices in the aquaculture sector, setting a new standard for technological applications in environmental and resource management. The advancements presented here provide a scalable and efficient solution for real-time monitoring, contributing to the long-term sustainability of marine ecosystems.

Laryngeal Cancer Screening During Flexible Video Laryngoscopy Using Large Computer Vision Models.

Develop an artificial intelligence assisted computer vision model to screen for laryngeal cancer during flexible laryngoscopy. Using laryngeal images and flexible laryngoscopy video recordings, we developed computer vision models to classify video frames for usability and cancer screening. A separate model segments any identified lesions on the frames. We used these computer vision models to construct a video stream annotation system. This system classifies findings from flexible laryngoscopy as "potentially malignant" or "probably benign" and segments any detected lesions. Additionally, the model provides a confidence level for each classification. The overall accuracy of the flexible laryngoscopy cancer screening model was 92%. For cancer screening, it achieved a sensitivity of 97.7% and a specificity of 76.9%. The segmentation model attained an average precision at a 0.50 intersection-over-union of 0.595. The confidence level for positive screening results can assist clinicians in counseling patients regarding the findings. Our model is highly sensitive and adequately specific for laryngeal cancer screening. Segmentation helps endoscopists identify and describe potential lesions. Further optimization is required to enable the model's deployment in clinical settings for real-time annotation during flexible laryngoscopy.