Down Syndrome Detection with CNN

I used followed methods

• A custom CNN model with 194,529 trainable parameters
• Splitted train(2549 images) and test set (450 images)
  
• Used elu activation function during feature extraction
• Used tf.data for input pipeline

CNN Model Architecture

Results

	Result
Test Accuracy	80.44 %
Test AUC	0.878
Test Precision	0.862
Test Recall	0.735

• Graphs

• Confusion Matrix