resnet-使用resnet+SVM实现乳腺癌检测-算法.zip

# Automatic Breast Cancer Detection From Hystopathology Images  - This is a Python project for breast cancer detection. The user has to give input of an `histopathological images` of `40x` magnification scale and the model will predict whether it is `cancerous (malignant)` or `non-cancerous (benign)`. - This entire end-to-end full-stack api utilizes popular frameworks like: - Flask - Jinja2 - OpenCV - Pytorch - Scikit-learn ## Description - The proposed framework is a hybrid of Deep Learning and traditional Machine Learning. - Following traditional `Transfer learning` technique `pre-trained ResNet18` model trained on BreakHis breast-cancer dataset is used for feature extraction. - High-dimensional `pre-final layer features` of ResNet18 having dimension of `512` are extracted and used for training and classification using the machine learning classifier `SVM`. - The overall flow diagram of the proposed method is shown in the following figure.  ## Dependencies pip install -r requirements.txt ## Folder structure Root folder |-----> static # The static folder containing unchangable frontend codes | |-----> css/index.css # CSS file for frontend | |-----> downloads # contains .png, .jpg, .jpeg files to be downloaded | |-----> js/index.js # JavaScript file for forntend | |-----> uploads # contains .png, .jpg, .jpeg files to be uploaded |-----> templates/index.html # the HTML file |-----> utils # Contains .py files and weights of trained ResNet18 | | (for features) and trained Support Vector Machine (SVM for classification) | |-----> classification # Contains the svm classifier model for training and prediction | | |-> train.py # Python code for training of SVM | | |-> svm_clf.py # Python code for prediction of cancer | |-----> ResNet18_features.py # Python code for features extraction from ResNet18 | |-----> model_ResNet18.pt # Trained weights of ResNet18 | |-----> svm_clf.pkl # Trained and saved SVM model | |-----> svm_clf_train_val.pkl # Model saved during validation (can be ignored) |-----> detector.py # Compunded Python file to perform the prediction |-----> app.py # The flask app to integrate frontend with backend