def view(label, image, seg_only, config):
'''View a spectrogram
This is a super function which provides all of the functionality to
view a spectrogram, either from the database itself or recreated
Args:
label: The label for the file
image: If not '', write image to a file named this
seg_only: View segments only
config: The parsed ini file for this particular run
Returns:
Nothing. It writes to MongoDB collection defined in the ini file
Raises:
FileNotFoundError: If the wavfile doesn't exist, it can't be processed
'''
# Get the data, from the database or recreate
init_client(config)
df, spectrogram, normalization_factor = read_spectrogram(label, config)
close_client()
# Apply Gaussian Filter
spectrogram = apply_gaussian_filter(
spectrogram, config['model_fit'].getfloat('gaussian_filter_sigma'))
# Either vizualize spectrogram w/ bounding boxes
# -> or, segments
if seg_only:
gen_segs(spectrogram, df, image)
else:
gen_spec_with_segs(label, spectrogram, df, image)
def chunk_run_prediction(chunk, train_labels_df, predict_labels_df, config):
'''Make a prediction based on a model
Given a directory of data, make a prediction using a model
Args:
chunk: A list of columns to process
train_labels_df: The training labels to run predictions from
predict_labels_df: The prediction labels to make predictions on
config: The parsed ini file for this run
Returns:
Nothing. Writes prediction data to MongoDB collection.
Raises:
A list of `model.predict_proba` outputs
'''
init_client(config)
predict_proba = [
run_prediction(col, train_labels_df, predict_labels_df, config)
for col in chunk
]
close_client()
return predict_proba
def chunk_preprocess(chunk, config):
'''Preprocess all images
This is a super function which provides all of the functionality to
preprocess many wav file for model fitting.
Args:
chunk: A chunk of files to process
config: The parsed ini file for this particular run
Returns:
Nothing
Raises:
Nothing
'''
# Before running the chunk, we need to initialize the global MongoDB client
init_client(config)
for label in chunk:
preprocess(label, config)
# Don't forget to close the client!
close_client()
def high_cc(l, species_found, config):
if len(l) != 0:
init_client(config)
all_file_file_statistics = generate_ff_stats(l, species_found)
close_client()
try:
results_str = ""
for idx, row in zip(l.index.values, all_file_file_statistics):
highest_cc = row.max()
if highest_cc >= 0.75:
build_str = np.array_str(row).replace('\n', '')
results_str += f"{idx}: {build_str}\n"
except ValueError:
pass
return results_str
else:
return ""
def chunk_run_stats(chunk, train_labels_df, config):
'''For each chunk call run_stats
Run within a parallel executor to generate file and file-file Statistics
for a given label.
Args:
chunk: A chunk of file labels
labels_df: Passed through to `file_file_stats` function
config: The parsed ini file for this run
Returns:
Nothing, writes to MongoDB
Raises:
Nothing.
'''
init_client(config)
[run_stats(label, train_labels_df, config) for label in chunk]
close_client()
def chunk_build_model(chunk, labels_df, config):
'''Build the lasseck2013 model
Given a chunk, run build_model on each label
Args:
chunk: Some columns to process in parallel
labels_df: The labels_df to build model with
config: The parsed ini file for this run
Returns:
Something or possibly writes to MongoDB
Raises:
Nothing.
'''
init_client(config)
results = [build_model(col, labels_df, config) for col in chunk]
close_client()
return results