This project involves classifying malaria-infected and uninfected cells using different machine learning and deep learning approaches. The primary goal is to automate the identification of parasitized and uninfected cells using microscopic images.
Malaria is a critical disease caused by parasites, and detecting it in cells is a time-consuming process for healthcare workers. This project aims to automate the process by using machine learning and deep learning models to classify cells as parasitized or uninfected from microscope images.
-
Performance:
- Accuracy on training data: ~65%
- Accuracy on test data: ~65%
While the model performs consistently on both train and test sets, the accuracy is too low to depend on for classification.
-
Performance:
- Accuracy on training data: 96%
- Accuracy on test data: 94%
The CNN model performs well with high accuracy on both training and test datasets.
-
Performance:
- Accuracy on training data: 96%
- Accuracy on test data: 94%
Adding regularization to the CNN model had no significant effect on the performance, resulting in similar accuracy to the base CNN model.
- Performance:
- Improved model performance, but hyperparameter tuning was time-consuming due to the large number of permutations tried. It was challenging to achieve faster training times with exhaustive tuning.
- Performance:
- Using transfer learning with the VGG19 architecture was applied for 1 epoch due to high computational cost.
- The model shows promise, but training for more epochs is needed for better performance.
Vivek Prakash