File: $ECE_8527/resources/data/set_15/lists/

This directory contains data extracted from the TNMG cardiology data.
Here is a brief description of the data.

In the lists directory, there are the following:

nedc_130_[1]: p
/data/isip/www/isip/courses/temple/ece_8527/resources/data/set_15
nedc_130_[1]: d
total 68
drwxrwxr-x  6 picone isip    8 Apr 28 20:03 ./
drwxrwxr-x 17 picone isip   18 Apr 14 02:22 ../
-rw-r--r--  1 picone isip 2803 Apr 28 20:03 AAREADME.txt
-r--r--r--  1 picone isip 3135 Apr 28 20:03 AAREADME.txt,v
drwxrwxr-x  3 picone isip    3 Apr 28 19:12 data/
drwxrwxr-x  3 picone isip    5 Apr 17 23:20 evaluation/
drwxrwxr-x  2 picone isip   22 Apr 28 19:58 lists/
drwx------  4 picone isip    6 Apr 19 01:06 subsets/

data/eval contains the eval data.

The lists directory includes these files:

nedc_130_[1]: ls -1 lists
...
data_train_healthy.csv
data_train_healthy.list
data_train_healthy.listh
data_train_unhealthy.csv
data_train_unhealthy.list
data_train_unhealthy.listh

data_dev_healthy.csv
data_dev_healthy.list
data_dev_healthy.listh
data_dev_unhealthy.csv
data_dev_unhealthy.list
data_dev_unhealthy.listh

data_eval.list
data_eval.listh

The filenames should be self-explanatory.

The data itself is stored in .dat files. These are data sampled at 300 Hz
and 2200 samples long. All files have the same length to make things easy.

This is 8-channel data: 1dAVb,RBBB,LBBB,SB,AF,ST. You do not need to worry
about what these mean initially. Just treat them as a single multichannel
signal so you model correlations between channels.

These .dat files contain 16-bit signed integers in an interleaved format:

 sig[0][0] - first sample, first channel
 sig[0][1] - first sample, second channel
 ...
 sig[0][7] - first sample, last channel
 sig[1][0] - second sample, first channel
 ...
 sig[1][7] - second sample, last channel
 ...
 sig[2199][0] - last sample, first channel
 ...
 sig[2199][7] - last sample, last channel, end of file
 
The annotations are stored in the corresponding csv files. These match the
*.list files line by line.

To run the scoring tool, try running this demo first:

nedc_130_[1]: p
/data/isip/www/isip/courses/temple/ece_8527/resources/data/set_15
nedc_130_[1]: sd evaluation/
/data/isip/www/isip/courses/temple/ece_8527/resources/data/set_15/evaluation
nedc_130_[1]: ./score.py tests/ref_tnmg.csv tests/hyp_tnmg.csv
Metric 1: simple accuracy
 err / acc = 0.0242 / 0.9758

Metric 2: micro accuracy / precision / recall / f1
 micro acc / prec / rec / f1 = 0.9960 / 0.9207 / 0.9557 / 0.9379

Metric 3: macro accuracy / precision / recall / f1
  [1dAVb] acc / prec / rec / f1 = 0.9927 / 0.8667 / 0.9286 / 0.8966
  [RBBB] acc / prec / rec / f1 = 0.9952 / 0.8947 / 1.0000 / 0.9444
  [LBBB] acc / prec / rec / f1 = 1.0000 / 1.0000 / 1.0000 / 1.0000
  [SB] acc / prec / rec / f1 = 0.9952 / 0.8333 / 0.9375 / 0.8824
  [AF] acc / prec / rec / f1 = 0.9964 / 1.0000 / 0.7692 / 0.8696
  [ST] acc / prec / rec / f1 = 0.9964 / 0.9474 / 0.9730 / 0.9600
 macro acc / prec / rec / f1 = 0.9960 / 0.9237 / 0.9347 / 0.9255

Your ref and hyp file should follow the format in these examples.