Pinned Repositories
A-proposed-model-that-can-predict-the-assessment-of-both-Syntax-Cohesion-Vocabulary-Phraseology-
The proposed model is able to predict the evaluation of both grammatical coherence, vocabulary and grammatical conventions, so that the evaluation can give each of those criteria a value between 1 and 5, I did not treat the system as a classification process, but rather it was treated as a REGRESSION issue. It includes several steps through which a few errors were reached, all ranging between 0.25 for each criterion. The values of the weights that were reached can also be used to deal with the issue as a classification process (but it was not dealt with as well in this proposed methodology).
Adult-tooth-segmentation-U-net-based-GAN-
Dental segmentation for adults. Many dentists find it difficult to analyze dental panoramic images for adults. One of the difficulties that dentists suffer from is the difficulty in determining the extent and root of the teeth, which affects the decisions of doctors in many cases that include dental implants, tooth extraction, or other problems.
Classification-of-dates-fruit-images-with-acc-99-
Suggesting a neural architecture for classifying dates images, using MobileNetV2 Transfer Learning
Convolutional-eXtreme-Gradient-Boosting-Brain-Tumor
Convolutional eXtreme Gradient Boosting for classification of brain tumors based on convolutional neural networks and XGBoost.
Documents-Classification-Using-CNN
Proposing the structure of a convolutional neural network, to identify the type of images of documents, the proposed structure of the convolutional neural network includes dividing the image of the document into four main regions and passing those regions to the convolutional neural network as a sequence of single images, and the goal of this methodology is the ability to facilitate the work of The convolutional neural network is able to extract the basic characteristics included in each of the four regions, and then use the GlobalAveragePooling1D layer in order to reach the general characteristics that distinguish the document, and thus the ability of the neural network to easily find the general characteristics that characterize each type of document.
Early-Detection-of-Collective-or-Individual-Theft-Attempts-Us-ing-Long-term-Recurrent-Convolutional-
I designed an intelligent system capable of analyzing movement within the videos and detecting suspicious movement that precedes the occurrence of shoplifting crimes. The proposed system can analyze the movement into two primary classifications: the natural movement, and the suspicious movement (with the percentage of each of them being determined.” Thus, the system appears, depending on the percentage of the type of movement, whether the possibility of theft is high or low, or the Confusion movement, which are branched cases depending on the percentage percent accuracy of smart model classification"). The system is integrated with surveillance camera systems that are placed in stores, and the system can at that time alert security personnel in cases where the movement of people in the monitored area appears to be suspicious. The system can also help in cases where it is required to search within a large number of video clips recorded by the surveillance cameras to determine the time moments before the theft crimes. The compressed file contains several video clips on which the system has been tested (the system is waiting for 160 frames to pass, “that is, approximately 3 seconds on average, depending on the frequency of the frames within the video clips or the live broadcast”). I sent you a detailed study of how the system works, and if you like the system and find that it can complement your software systems, I will send you the code and the smart trained model.
Human-Protein-Atlas-Image-Classification
Proposing a neural network architecture capable of classifying protein organelle localization labels, the proposed model was able to reach an accuracy of 95 percent for test data and training data. The proposed model deals with the input of the proposed neural network as three-dimensional (each dimension represents colors (red filter, blue filter, yellow filter, green filter)), and thus the input to the neural network (samples, 2, 90, 90, 3) represents the number 2 In the proposed structure (RGB image, image with yellow filter).
Multi-Scale-CycleGAN-Night-to-Day
Converting night into day is one of the most interesting applications in generative models, due to the great difficulty in recreating the scene during the day, especially in cases of extreme darkness, and thus the difficulty lies in imagining the scene during the day when the lighting is very weak.
Simple-Chatbot
Chatbot using RNN with Embedding Transfer Learning
VAE-CycleGAN-MRI-CT-Scan-Images
The study works on generating CT images from MRI images, where unsupervised learning was used using VAE-CycleGan. Since the number of samples included in the data set used in the study, and therefore in this case we are in a state of epistemic uncertainty, therefore probabilistic models were used in forming the latent space.
kaledhoshme123's Repositories
kaledhoshme123/Early-Detection-of-Collective-or-Individual-Theft-Attempts-Us-ing-Long-term-Recurrent-Convolutional-
I designed an intelligent system capable of analyzing movement within the videos and detecting suspicious movement that precedes the occurrence of shoplifting crimes. The proposed system can analyze the movement into two primary classifications: the natural movement, and the suspicious movement (with the percentage of each of them being determined.” Thus, the system appears, depending on the percentage of the type of movement, whether the possibility of theft is high or low, or the Confusion movement, which are branched cases depending on the percentage percent accuracy of smart model classification"). The system is integrated with surveillance camera systems that are placed in stores, and the system can at that time alert security personnel in cases where the movement of people in the monitored area appears to be suspicious. The system can also help in cases where it is required to search within a large number of video clips recorded by the surveillance cameras to determine the time moments before the theft crimes. The compressed file contains several video clips on which the system has been tested (the system is waiting for 160 frames to pass, “that is, approximately 3 seconds on average, depending on the frequency of the frames within the video clips or the live broadcast”). I sent you a detailed study of how the system works, and if you like the system and find that it can complement your software systems, I will send you the code and the smart trained model.
kaledhoshme123/Multi-Scale-CycleGAN-Night-to-Day
Converting night into day is one of the most interesting applications in generative models, due to the great difficulty in recreating the scene during the day, especially in cases of extreme darkness, and thus the difficulty lies in imagining the scene during the day when the lighting is very weak.
kaledhoshme123/VAE-CycleGAN-MRI-CT-Scan-Images
The study works on generating CT images from MRI images, where unsupervised learning was used using VAE-CycleGan. Since the number of samples included in the data set used in the study, and therefore in this case we are in a state of epistemic uncertainty, therefore probabilistic models were used in forming the latent space.
kaledhoshme123/Human-Protein-Atlas-Image-Classification
Proposing a neural network architecture capable of classifying protein organelle localization labels, the proposed model was able to reach an accuracy of 95 percent for test data and training data. The proposed model deals with the input of the proposed neural network as three-dimensional (each dimension represents colors (red filter, blue filter, yellow filter, green filter)), and thus the input to the neural network (samples, 2, 90, 90, 3) represents the number 2 In the proposed structure (RGB image, image with yellow filter).
kaledhoshme123/Convolutional-eXtreme-Gradient-Boosting-Brain-Tumor
Convolutional eXtreme Gradient Boosting for classification of brain tumors based on convolutional neural networks and XGBoost.
kaledhoshme123/Documents-Classification-Using-CNN
Proposing the structure of a convolutional neural network, to identify the type of images of documents, the proposed structure of the convolutional neural network includes dividing the image of the document into four main regions and passing those regions to the convolutional neural network as a sequence of single images, and the goal of this methodology is the ability to facilitate the work of The convolutional neural network is able to extract the basic characteristics included in each of the four regions, and then use the GlobalAveragePooling1D layer in order to reach the general characteristics that distinguish the document, and thus the ability of the neural network to easily find the general characteristics that characterize each type of document.
kaledhoshme123/A-proposed-model-that-can-predict-the-assessment-of-both-Syntax-Cohesion-Vocabulary-Phraseology-
The proposed model is able to predict the evaluation of both grammatical coherence, vocabulary and grammatical conventions, so that the evaluation can give each of those criteria a value between 1 and 5, I did not treat the system as a classification process, but rather it was treated as a REGRESSION issue. It includes several steps through which a few errors were reached, all ranging between 0.25 for each criterion. The values of the weights that were reached can also be used to deal with the issue as a classification process (but it was not dealt with as well in this proposed methodology).
kaledhoshme123/Adult-tooth-segmentation-U-net-based-GAN-
Dental segmentation for adults. Many dentists find it difficult to analyze dental panoramic images for adults. One of the difficulties that dentists suffer from is the difficulty in determining the extent and root of the teeth, which affects the decisions of doctors in many cases that include dental implants, tooth extraction, or other problems.
kaledhoshme123/Semantic-Similarity-using-TimeDistributed-LSTM
The following notebook, reviews the methodology by which we can build a recurrent neural network that is able to analyze text sentences and determine whether they are congruent in meaning or contradictory in terms of meaning
kaledhoshme123/Transformers-Text-Classification
Suggesting a neural network architecture for analyzing and recognizing texts, where transformers were used through a pre-trained BERT model, in addition to its integration with the LSTM layer with the Global Pooling layers, in order to reach a model capable of analyzing texts.
kaledhoshme123/Using-GAN-to-Generate-Chest-X-Ray-Images
The following study presents a model for generating chest X-ray images of normal subjects (without lung disease) and pneumonia patients.
kaledhoshme123/Analysis-and-sorting-of-the-feelings-of-the-tweeters-regarding-the-events-in-Sri-Lanka
The following code illustrates the mechanism by which we can aggregate Tweets based on sentiment. The aggregation process is based on the association of tweets with the same feelings, as well as the degree and proportion of the feeling. The methodology used is based on building a recurrent neural network capable of analyzing sentiment, using a data set that includes a number of emotions. The next stage involves using the trained model to sort tweets based on sentiment with a rating ratio. In this partial stage, we will follow two methodologies: The first is to draw a graph that shows the percentage of each of the feelings of the tweeters within Twitter regarding what is happening in the state of Sri Lanka. The next partial stage, is to move to the study of each of these feelings for the tweeters, and try to collect them in order to determine the degree of feelings for each of them. The final hierarchical schemas (for each one of the feelings) will show the correlation of the tweeters in terms of the degree of affiliation with that feeling. The Euclidean distance was used to calculate the degree of convergence for a single feeling (depending on the percentage of tweeting classification and belonging to a specific feeling).
kaledhoshme123/Breast-cancer-segmentation-malignant-benign-normal-
Breast cancer is one of the most common causes of death among women worldwide. Early detection helps reduce the number of premature deaths. In the study, I am working on creating a convolutional neural network capable of identifying tumor areas within medical images (which were taken with ultrasound).
kaledhoshme123/Classification-foliar-diseases-in-apple-trees-with-accuracy-99-
Misdiagnosis of many diseases affecting agricultural crops can lead to chemical misuse resulting in the emergence of resistant pathogenic strains, increased input costs, and more outbreaks leading to significant economic losses and environmental impacts. A structure of a convolutional neural network has been proposed that is capable of diagnosing disease in apple plant leaves. The proposed neural network structure was able to reach an accuracy of more than 99 percent.
kaledhoshme123/Is-it-similar-to-a-previous-medical-condition
In many cases that doctors face during the different treatment processes for many patients, so that the doctor tries every time to remember an old pathological condition that he encountered in advance in order to try to retrieve the methodology that he followed at that moment to treat the patient who has the same pathological condition.
kaledhoshme123/Merging-styles-two-face-images-at-middle-details-
In this study, we work to transfer the pattern to the basic face image from the pattern face image, by adjusting the latent space and moving within it.
kaledhoshme123/Pizza-or-Not-Pizza-using-VGG19-Transfer-Learning
Suggesting a convolutional neural network to identify pizza and distinguish it from other diets
kaledhoshme123/Predict-age-of-patient-based-on-the-X-Ray-images
create a model capable of predicting the patient's age group through chest X-rays.
kaledhoshme123/Transformers-LSTM-with-XGBRegressor
The proposed model is able to predict the evaluation of both grammatical coherence, vocabulary and grammatical conventions, so that the evaluation can give each of those criteria a value between 1 and 5, I did not treat the system as a classification process, but rather it was treated as a REGRESSION issue
kaledhoshme123/Use-Conditional-Variational-Autoencoders-to-extract-important-information
The study relied on conditional Variational Autoencoders to generate x-ray images, so that we can be able to regenerate the images according to the most important information that the x-ray images can contain (important information extraction).
kaledhoshme123/Using-Deep-Learning-to-Annotate-the-Protein-Universe
Understanding the relationship between amino acid sequence and protein function is a long-standing problem in molecular biology with far-reaching scientific implications.
kaledhoshme123/X-ray-Covid-19---Pneumonia-Heat-map
Obtaining a Heat Map of the areas most influential in sorting chest X-ray images.
kaledhoshme123/Probabilistic-U-Net-Segmentation-Ambiguous-Images-
People with pulmonary disease often have a high opacity, which makes segmentation of the lung from chest X-rays more difficult. In this study, I propose a methodology to improve the performance of the U-NET structure so that it is able to extract the features and spatial characteristics of the X-ray images of the chest region.
kaledhoshme123/Determine-the-location-of-the-skin-lesion-spread
Dermatologists suffer from the difficulty of locating cancerous and malignant skin lesions, which causes many problems during the process of removing the tumor, which leads to the return of the tumor again. In determining the location of the tumor and its spread and determining the area that must be removed accurately.
kaledhoshme123/Digit-Recognizer-CNN-acc-100-
identify digits from a dataset of tens of thousands of handwritten images
kaledhoshme123/Generate-semi-conditional-data-augmentation-X-ray-Images
The following study, through which we can generate X-ray images of the chest region in a semi-conditional manner, by taking advantage of the probability distributions.
kaledhoshme123/Google-Brain---Ventilator-Pressure-Prediction
Simulate a ventilator connected to a sedated patient's lung
kaledhoshme123/kaledhoshme123
kaledhoshme123/Multimodal-face-generation-facial-biometrics-
Similarity between faces: One person resembles another person to a large degree. This can lead to many problems facing security surveillance systems. Facial recognition systems have difficulty distinguishing between the main person and other people who are highly similar in terms of features.
kaledhoshme123/Semantic-Segmentation-of-the-lungs-from-X-ray-images
A model of the segmentation of the lung region from the x-ray images, the model relies on forming a mask through which that region can be cut.