def add_exports(self):
# on MacOS we will not use os.fork, elsewhere this does nothing.
forking_enable(0)
outfiles = set()
for d in self.new_exports:
logger.debug("Adding new export.")
should_terminate = Value(c_bool, False)
frames_to_export = Value(c_int, 0)
current_frame = Value(c_int, 0)
start_frame = None
end_frame = None
export_dir = d
user_dir = self.g_pool.user_dir
#we need to know the timestamps of our exports.
try: # 0.4
frames_to_export.value = len(
np.load(os.path.join(export_dir, 'world_timestamps.npy')))
except: # <0.4
frames_to_export.value = len(
np.load(os.path.join(export_dir, 'timestamps.npy')))
# Here we make clones of every plugin that supports it.
# So it runs in the current config when we lauch the exporter.
plugins = self.g_pool.plugins.get_initializers()
#make a unique name created from rec_session and dir name
rec_session, rec_dir = export_dir.rsplit(os.path.sep, 2)[1:]
out_name = rec_session + "_" + rec_dir + ".mp4"
out_file_path = os.path.join(self.destination_dir, out_name)
if out_file_path in outfiles:
logger.error(
"This export setting would try to save %s at least twice please rename dirs to prevent this. Skipping File"
% out_file_path)
else:
outfiles.add(out_file_path)
logger.info("Exporting to: %s" % out_file_path)
process = Export_Process(
target=export,
args=(should_terminate, frames_to_export, current_frame,
export_dir, user_dir,
self.g_pool.min_data_confidence, start_frame,
end_frame, plugins, out_file_path))
self.exports.append(process)
def main():
import argparse
from textwrap import dedent
from file_methods import Persistent_Dict
def show_progess(jobs):
no_jobs = len(jobs)
width = 80
full = width / no_jobs
string = ""
for j in jobs:
try:
p = int(width * j.current_frame.value /
float(j.frames_to_export.value * no_jobs))
except:
p = 0
string += '[' + p * "|" + (full - p) * "-" + "]"
sys.stdout.write("\r" + string)
sys.stdout.flush()
"""Batch process recordings to produce visualizations
Using simple_circle as the default visualizations
Steps:
- User Supplies: Directory that contains many recording(s) dirs or just one recordings dir
- We walk the user supplied directory to get all data folders
- Data is the list we feed to our multiprocessed
- Error check -- do we have required files in each dir?: world.avi, gaze_positions.npy, timestamps.npy
- Result: world_viz.avi within each original data folder
"""
parser = argparse.ArgumentParser(
formatter_class=argparse.RawDescriptionHelpFormatter,
description=dedent('''\
***************************************************
Batch process recordings to produce visualizations
The default visualization will use simple_circle
Usage Example:
python batch_exporter.py -d /path/to/folder-with-many-recordings -s ~/Pupil_Player/settings/user_settings -e ~/my_export_dir
Arguments:
-d : Specify a recording directory.
This could have one or many recordings contained within it.
We will recurse into the dir.
-s : Specify path to Pupil Player user_settings file to use last used vizualization settings.
-e : Specify export directory if you dont want the export saved within each recording dir.
-p : Export a 120 frame preview only.
***************************************************\
'''))
parser.add_argument('-d', '--rec-dir', required=True)
parser.add_argument('-s', '--settings-file', required=True)
parser.add_argument('-e', '--export-to-dir', default=False)
parser.add_argument('-c', '--basic-color', default='red')
parser.add_argument('-p', '--preview', action='store_true')
if len(sys.argv) == 1:
print(parser.description)
return
args = parser.parse_args()
# get the top level data folder from terminal argument
data_dir = args.rec_dir
if args.settings_file and os.path.isfile(args.settings_file):
session_settings = Persistent_Dict(
os.path.splitext(args.settings_file)[0])
# these are loaded based on user settings
plugin_initializers = session_settings.get('loaded_plugins', [])
session_settings.close()
else:
logger.error("Setting file not found or valid")
return
if args.export_to_dir:
export_dir = args.export_to_dir
if os.path.isdir(export_dir):
logger.info("Exporting all vids to {}".format(export_dir))
else:
logger.error("Exporting dir is not valid {}".format(export_dir))
return
else:
export_dir = None
logger.info("Exporting into the recording dirs.")
if args.preview:
preview = True
logger.info("Exporting first 120frames only")
else:
preview = False
class Temp(object):
pass
recording_dirs = get_recording_dirs(data_dir)
# start multiprocessing engine
n_cpu = mp.cpu_count()
logger.info(
"Using a maximum of {} CPUs to process visualizations in parallel...".
format(n_cpu))
jobs = []
outfiles = set()
for d in recording_dirs:
j = Temp()
logger.info("Adding new export: {}".format(d))
j.should_terminate = mp.Value(c_bool, 0)
j.frames_to_export = mp.Value(c_int, 0)
j.current_frame = mp.Value(c_int, 0)
j.data_dir = d
j.user_dir = None
j.start_frame = None
if preview:
j.end_frame = 30
else:
j.end_frame = None
j.plugin_initializers = plugin_initializers[:]
if export_dir:
# make a unique name created from rec_session and dir name
rec_session, rec_dir = d.rsplit(os.path.sep, 2)[1:]
out_name = rec_session + "_" + rec_dir + ".mp4"
j.out_file_path = os.path.join(os.path.expanduser(export_dir),
out_name)
if j.out_file_path in outfiles:
logger.error(
"This export setting would try to save {} at least twice pleace rename dirs to prevent this."
.format(j.out_file_path))
return
outfiles.add(j.out_file_path)
logger.info("Exporting to: {}".format(j.out_file_path))
else:
j.out_file_path = None
j.args = (j.should_terminate, j.frames_to_export, j.current_frame,
j.data_dir, j.user_dir, j.start_frame, j.end_frame,
j.plugin_initializers, j.out_file_path, None)
jobs.append(j)
todo = jobs[:]
workers = [
Export_Process(target=export, args=todo.pop(0).args)
for i in range(min(len(todo), n_cpu))
]
for w in workers:
w.start()
working = True
while working: # cannot use pool as it does not allow shared memory
working = False
for i in range(len(workers)):
if workers[i].is_alive():
working = True
else:
if todo:
workers[i] = Export_Process(target=export,
args=todo.pop(0).args)
workers[i].start()
working = True
show_progess(jobs)
time.sleep(.25)
print('\n')