From 2d51e275f2e6c9897335079a65f78e3f70a7a2aa Mon Sep 17 00:00:00 2001 From: roboswell Date: Sun, 23 Jun 2024 00:52:58 +0800 Subject: [PATCH] Revised README file --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index 75a497f..b752a9a 100644 --- a/README.md +++ b/README.md @@ -40,5 +40,5 @@ - CNN Deep Learning for Medical Imaging: - **Title:** ["Comparative Analysis of CNN Deep Learning Models for X-ray Illness Classification"](./Neural%20Network%20Models/X-Ray%20Deep%20Learning%20Classificaton%20Models.html) - Focus: Deep Learning for Image Classification - Technology Used: Python, Keras, CNNs, Transfer Learning, ImageDataGenerator, flow_from_directory, EarlyStopping, ModelCheckpoint, ReduceLROnPlateau - - Contents: Comparing the performance of 8 CNN deep learning models on X-ray images from three classes (COVID-19, viral pneumonia, and healthy). These include transfer learning models (e.g., InceptionV3), and various techniques to improve model generalization and help avoid overfitting (e.g., dropout, batch normalization, early stopping, data augmentation, L1 and L2 regularization, fire modules, and ways of using deep networks effectively). I also demonstrate best practices for structuring filters/kernels, channels, layers, activation functions, pooling, convolutional blocks, and other model components for optimal performance. Metrics include confusion matrixes, accuracy, precision, recall, F1-score, ROC curve, and AUC. Analysis of non-augmented vs. augmented data models with specific augmentation techniques are shown. Architectures and training strategies for each model are detailed. \ No newline at end of file + - Technology Used: Python, Keras, CNNs, Transfer Learning, ImageDataGenerator, flow_from_directory, EarlyStopping, ModelCheckpoint, ReduceLROnPlateau + - Contents: Comparing the performance of 8 CNN deep learning models on X-ray images from three classes (COVID-19, viral pneumonia, and healthy). These include transfer learning models (e.g., InceptionV3), and various techniques to improve model generalization and help avoid overfitting (e.g., dropout, batch normalization, early stopping, data augmentation, L1 and L2 regularization, fire modules, and ways of using deep networks effectively). I also demonstrate best practices for structuring filters/kernels, channels, layers, activation functions, pooling, convolutional blocks, and other model components for optimal performance. Metrics include confusion matrixes, accuracy, precision, recall, F1-score, ROC curve, and AUC. Analysis of non-augmented vs. augmented data models with specific augmentation techniques are shown. Architectures and training strategies for each model are detailed. \ No newline at end of file