ECG Time Series Classification

Fine-tuning VGG16 and Inception-Resnet V2 to classify ECG signals in two different classes, basically normal and abnormal.

Data

1. 100 original TRAIN data, and 100 original TEST data. Here is some samples from TRAIN set:
2. Split TRAIN into 90 train set and 10 val(validation) set;
3. Augment train and val sets:
   • train: (1440, 96)
   • val: (160, 96)

VGG16

Results

• Achieve 0.955 TEST accuracy for now on TEST dataset with 100 data
• train acc and loss:

• validation acc:

• Evaluate TEST set: