E
x
C
han
G
eAI
Analyse
Predict
Finetune
Data
Sampling rate:
100hz
500hz
Custom
ADC gain:
200/mV
1000/mV
Lead layout (I-III ... V1-V6):
aVR, aVL, aVF
aVR, aVF, aVL
Upload Labels (optional)
Model
E
x
C
han
G
eAI
Prediction
Training
E
x
C
han
G
eAI
Load external Models
Cancel
Save
Summary
imported ECGs
imported Labels
Clear loaded Data
Export
Based on:
Raw Time Series
QRS-Complex
Rlign
Rlign-MedianBeats
Events
json
csv
Export ECGs
E
x
C
han
G
eAI
This page is not optimized for your device size.
Finetuning
Select AI model:
Training Method
?
:
Finetuning (all layers)
Trains all layers
Leverages Feature extraction from base model, but can finetune this process as well
Slower, but more accurate
Finetuning (classification head)
Trains only the top layers (classification heads)
Based on Feature extraction from base model
Faster, but less accurate and depends on the extracted features
Finetuning (classification head)
Finetuning (all layers)
Custom model name:
Advanced Settings
Optimizer:
AdamW
Adam
NAdam
Adamax
Adafactor
SGD
RMSprop
Maximal initial learning-rate:
Batch size:
Number of epochs:
LR-scheduler gamma:
Show logs:
Dataset Analytics:
Number of imported ECGs:
Number of imported Labels:
Start finetuning