testscript_deterministicwithResNet152.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
#
# DeepLabCut Toolbox (deeplabcut.org)
# © A. & M.W. Mathis Labs
# https://github.com/DeepLabCut/DeepLabCut
#
# Please see AUTHORS for contributors.
# https://github.com/DeepLabCut/DeepLabCut/blob/master/AUTHORS
#
# Licensed under GNU Lesser General Public License v3.0
#
"""
Created on Tue Oct  2 13:56:11 2018
@author: alex

DEVELOPERS:
This script tests various functionalities in an automatic way.

Note that the same ResNet 152 is trained 4 times (twice with the standard loader).
The sequence of losses is different....

Then twice with the 'deterministic' loader. The losses are identical when this is done twice.

I.e. I get twice: ;)
iteration: 1 loss: 1.6505 lr: 0.001
iteration: 2 loss: 0.6929 lr: 0.001
iteration: 3 loss: 0.6420 lr: 0.001
iteration: 4 loss: 0.5579 lr: 0.001
iteration: 5 loss: 0.4746 lr: 0.001
iteration: 6 loss: 0.3366 lr: 0.001
iteration: 7 loss: 0.3194 lr: 0.001
iteration: 8 loss: 0.2561 lr: 0.001
iteration: 9 loss: 0.1964 lr: 0.001
iteration: 10 loss: 0.1220 lr: 0.001


It produces nothing of interest scientifically.
"""

task = "TEST-deterministic"  # Enter the name of your experiment Task
scorer = "Alex"  # Enter the name of the experimenter/labeler


import os, subprocess, deeplabcut
from pathlib import Path
import pandas as pd
import numpy as np

print("Imported DLC!")
basepath = os.path.dirname(os.path.abspath("testscript.py"))
videoname = "reachingvideo1"
video = [
    os.path.join(
        basepath, "Reaching-Mackenzie-2018-08-30", "videos", videoname + ".avi"
    )
]

# to test destination folder:
# dfolder=basepath
dfolder = None

print("CREATING PROJECT")
path_config_file = deeplabcut.create_new_project(task, scorer, video, copy_videos=True)

cfg = deeplabcut.auxiliaryfunctions.read_config(path_config_file)
cfg["numframes2pick"] = 5
cfg["pcutoff"] = 0.01
cfg["TrainingFraction"] = [0.8]
cfg["default_net_type"] = "resnet_152"  #'mobilenet_v2_0.35'

deeplabcut.auxiliaryfunctions.write_config(path_config_file, cfg)

print("EXTRACTING FRAMES")
deeplabcut.extract_frames(path_config_file, mode="automatic", userfeedback=False)

print("CREATING-SOME LABELS FOR THE FRAMES")
frames = os.listdir(os.path.join(cfg["project_path"], "labeled-data", videoname))
# As this next step is manual, we update the labels by putting them on the diagonal (fixed for all frames)
for index, bodypart in enumerate(cfg["bodyparts"]):
    columnindex = pd.MultiIndex.from_product(
        [[scorer], [bodypart], ["x", "y"]], names=["scorer", "bodyparts", "coords"]
    )
    frame = pd.DataFrame(
        100 + np.ones((len(frames), 2)) * 50 * index,
        columns=columnindex,
        index=[os.path.join("labeled-data", videoname, fn) for fn in frames],
    )
    if index == 0:
        dataFrame = frame
    else:
        dataFrame = pd.concat([dataFrame, frame], axis=1)

dataFrame.to_csv(
    os.path.join(
        cfg["project_path"],
        "labeled-data",
        videoname,
        "CollectedData_" + scorer + ".csv",
    )
)
dataFrame.to_hdf(
    os.path.join(
        cfg["project_path"],
        "labeled-data",
        videoname,
        "CollectedData_" + scorer + ".h5",
    ),
    "df_with_missing",
    format="table",
    mode="w",
)

print("Plot labels...")

deeplabcut.check_labels(path_config_file)

print("CREATING TRAININGSET")
deeplabcut.create_training_dataset(path_config_file)

# posefile=os.path.join(cfg['project_path'],'dlc-models/iteration-'+str(cfg['iteration'])+'/'+ cfg['Task'] + cfg['date'] + '-trainset' + str(int(cfg['TrainingFraction'][0] * 100)) + 'shuffle' + str(1),'train/pose_cfg.yaml')

shuffle = 1
posefile, _, _ = deeplabcut.return_train_network_path(path_config_file, shuffle=shuffle)
print("CHANGING training parameters to end quickly!")
edits = {"save_iters": 4, "display_iters": 1, "multi_step": [[0.001, 10]]}
DLC_config = deeplabcut.auxiliaryfunctions.edit_config(posefile, edits)

print("TRAIN")
deeplabcut.train_network(path_config_file)
print("TRAIN again... different loss?")
deeplabcut.train_network(path_config_file)

DLC_config = deeplabcut.auxiliaryfunctions.edit_config(
    posefile, {"dataset_type": "deterministic", "deterministic": True}
)

print("TRAIN")
deeplabcut.train_network(path_config_file)

print("TRAIN again... the same losses!")
deeplabcut.train_network(path_config_file)

print("ALL DONE!!! - deterministic at least runs... were the losses identical?")