# file: $NEDC_NFC/util/python/nedc_eeg_resnet/src/nedc_eeg_resnet_decode/params_v00.txt
#
# A parameter file that defines key paramters for decoding EEG signals.
# The standard NEDC parameter file parser is used to load this. There is
# a function in the decoder class, called load_parameters, that loads this.
#
# revision history:
#
# 20240118 (DH): converted to toml
# 20240923 (DH): replaced channels property with a montage path
# 20220818 (JP): another code review
# 20220816 (JP): code review
# 20220721 (ML): initial version
#
# set version information
#
version = "param_v1.0.0"

#------------------------------------------------------------------------------
#

[RESNET_DECODE]

# sample_frequency is measured in Hz (50 Hz).
# frame_duration is the duration of a frame in secs (256 samples @ 50 Hz).
# transform_crop is the number of samples we throw away (32 samples).
# The transform order is frame_duration_in_samples - transform_crop.
#
sample_frequency = 50.000
frame_duration = 5.1200
transform_crop = 32

# Transforms are different for each class
# Notes:
#  Using Imagenet data is common practice when using images in deep
#   machine learning, this is because of the large database that ImageNet
#   boasts, (over 1 million classified images) because if this their values
#   for mean, standard deviation, resize and crop tend to do very well
#   and all of these values are taken from the ImageNet database, more
#   information about ImageNet can be found here:
#   https://www.image-net.org/about.php	
#	
#  transforms.Resize resizes the image to a specified length and width
#  transforms.CenterCrop crops the image to a specified length and width
#  transforms.ToTensor converts the PIL image to tensor format and scales
#    the pixel values between 0.0 and 1.0
#  transforms.Normalize normalizes (smooths) the image by applying the
#    formula: out = (in - mean) / standard deviation, the first list
#    contains mean values for red, green, and blue channels (an RGB image has
#    3 channels, red, green, and blue), the second list contains values for
#    the standard deviation also for the red, green, and blue channels
#
transforms = 'transforms.Compose([transforms.Resize(frmsize), transforms.CenterCrop(transform_order), transforms.ToTensor(), transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])])'


# define the number of batched frames the data loader will use
# for evaluating the signals with the trained model
# generally the larger the batch size the faster but requires
# more memory
#
batch_size = 20

# define the number of workers we want to use for the data loader
# num workers will make make I/O operations faster
#
num_workers = 2

# define the number of threads we want to use (only for use of CPU)
# number of threads will make large tensor operations faster
#
num_threads = 16

# montage property:
#  
montage = "$NEDC_NFC/util/python/nedc_eeg_resnet/v2.0.1/montage/montage_resnet_v00.txt"

# model to decode on
#
model_path = "$NEDC_NFC/util/python/nedc_eeg_resnet/v2.0.1/models/model.pckl"

# pre-trained model path
#
pre_trained_model_path = "$NEDC_NFC/util/python/nedc_eeg_resnet/v2.0.1/models/resnet-18_pretrained.pth"
    
# select a CPU or GPU device:
#  decoding does not run significantly faster on a GPU, so we use
#  CPU's for decoding since we have a lot of available CPU's.
#
device = "cpu"

# set the post processor parameters:
#  seizure_threshold: the minimum probability for a seizure event
#  minimum_background_duration: the minimum time for an acceptable
#     background event
#  minimum_seizure_duration: the minimum time for an acceptable
#     seizure event
#
seizure_threshold = 0.90
minimum_background_duration = 40
minimum_seizure_duration = 20

#------------------------------------------------------------------------------
#
# end of file
#
#------------------------------------------------------------------------------