CNN-BloodMNIST

Training a CNN on the BloodMNIST dataset for effective categorisation of bloodcell types

Highest achieved accuracy so far on validation set with small cpu-trainable model: 95%

Target (literature documented transfer-learning result): 98%