search

sicara / easy-few-shot-learning

Ready-to-use code and tutorial notebooks to boost your way into few-shot learning for image classification.
MIT License
1.03k stars 141 forks source link

Can I use it for ECG class classification? #156

Open charlielol8 opened 1 month ago

charlielol8 commented 1 month ago

Problem I try to do a Ventricular Fibrillation (VF) and ventricular tachycardia (VT) using The MIT-BIH Malignant Ventricular Arrhythmia Database for few-shot learning where I want to segment it into windows when the label of interest is detected and put them into the data frame and split it to support and query set for training and testing.

I am new to this field, so i was wondering if this type of classification work in Few shot scenario and if so what kind of ecg files should I work with? For example should it be .npy, .csv, or image files like png or jpg? Also how can I split the data to into support and query set for training and testing

Could you maybe suggest on the insight on how i could do this project.

ebennequin commented 1 month ago

Hi. I don't have any experience working with ECG files so I cannot help you with that. Interfaces in easyfsl are pretty generic and should apply to different kinds of source data, but you probably need some extra layers. Feel free to share your progress!