MILCA

MULTI-INSTANCE LEARNING for Malaria Detection

MIL Script

python mil_vgg19_random_malaria.py \
  --dataset /path/to/your/dataset1 \
  --dataset2 /path/to/your/dataset2 \
  --csv_labels /path/to/your/train_labels.csv \
  --test_csv_labels /path/to/your/test_labels.csv \
  --output_dir /path/to/your/output_dir \
  --save_dir /path/to/your/save_dir \
  --num_labels 2 \
  --imsize 512 \
  --crop_size 512 \
  --num_iterations 10001 \
  --pretrained_model /path/to/your/pretrained_model.npy

Argument	Description
--dataset	Path to the primary dataset directory containing image tiles for training.
--dataset2	Path to a secondary dataset directory (optional, used to combine datasets).
--csv_labels	CSV file containing image labels for training.
--test_csv_labels	CSV file containing image labels for testing/validation.
--output_dir	Directory where training outputs, logs, and results will be saved.
--save_dir	Directory where trained model checkpoints will be stored.
--num_labels	Number of output labels (classes) for the classification task.
--imsize	Image resize dimension before cropping (in pixels).
--crop_size	Final cropped image size used during training (in pixels).
--num_iterations	Number of training iterations.
--pretrained_model	Path to the pretrained model weights (.npy file) used to initialize the network.