Supp. Figure 2 - Examples of WNet3D losses, and output during training#

  • Goal : show the output of the WNet3D during training, and the corresponding losses

  • The take_output_screenshots notebook was used to generate the screenshots during training

import pandas as pd
import sys
sys.path.append("../..")

from utils import *
from plots import *
show_params()
#################
SAVE_PLOTS_AS_PNG = False
SAVE_PLOTS_AS_SVG = True

Warning: Saturation in 0.9508196721311476 is too low or too high in hex color 7209B7
Plot parameters (set in plots.py) : 
- COLORMAP : ████████
- DPI : 200
- Data path : C:\Users\Cyril\Desktop\Code\CELLSEG_BENCHMARK
- Font size : 20
- Title font size : 25.0
- Label font size : 20.0

%load_ext autoreload
%autoreload 2

Loss values#

csv_path = "../../other/fig3-x-losses.csv"
if not Path(csv_path).is_file():
    log_path = "../../other/log.txt"
    losses, epochs = extract_losses_from_log(log_path)
    losses_df = pd.DataFrame.from_dict(losses, orient='index')

    # save as csv in other
    losses_df.to_csv("../../other/fig3-x-losses.csv")
else:
    losses_df = pd.read_csv(csv_path, index_col=0)

losses_df.index.name = 'Epoch'
losses_df.columns = ["SoftNCuts loss", "Reconstruction loss", "Weighted sum of losses"]
losses_df
SoftNCuts loss Reconstruction loss Weighted sum of losses
Epoch
1 0.99227 57.08382 29.03805
2 0.99046 45.87604 23.43325
3 0.92455 42.65216 21.78835
4 0.75902 40.94435 20.85169
5 0.66575 39.98960 20.32768
6 0.60914 39.26025 19.93469
7 0.58295 38.63013 19.60654
8 0.58262 38.05232 19.31747
9 0.56323 37.59191 19.07757
10 0.54814 37.17325 18.86069
11 0.52946 36.81464 18.67205
12 0.52598 36.46110 18.49354
13 0.51152 36.14809 18.32981
14 0.49954 35.84817 18.17385
15 0.48920 35.58164 18.03542
16 0.47675 35.31505 17.89590
17 0.46733 35.06290 17.76511
18 0.45689 34.82457 17.64073
19 0.44603 34.59145 17.51874
20 0.43645 34.34674 17.39160
21 0.42316 34.11471 17.26893
22 0.41604 33.90068 17.15836
23 0.40337 33.66850 17.03593
24 0.39882 33.46028 16.92955
25 0.38790 33.24519 16.81654
26 0.37429 33.03520 16.70475
27 0.37117 32.82695 16.59906
28 0.36206 32.62820 16.49513
29 0.35596 32.40600 16.38098
30 0.34413 32.19054 16.26733
31 0.33634 32.00090 16.16862
32 0.33110 31.80333 16.06721
33 0.32767 31.58601 15.95684
34 0.32160 31.38450 15.85305
35 0.31506 31.19447 15.75477
36 0.30943 30.99892 15.65417
37 0.30781 30.79648 15.55215
38 0.29929 30.62829 15.46379
39 0.29920 30.42325 15.36122
40 0.28847 30.21615 15.25231
41 0.28418 30.03862 15.16140
42 0.27881 29.85732 15.06807
43 0.27991 29.66614 14.97303
44 0.27281 29.46947 14.87114
45 0.26895 29.29379 14.78137
46 0.27796 29.12376 14.70086
47 0.26320 28.94697 14.60509
48 0.25966 28.74820 14.50393
49 0.25830 28.56196 14.41013
50 0.24921 28.38370 14.31645

Plots#

plot_losses(losses_df, losses_df.columns, colormap=[COLORMAP[1], COLORMAP[3], COLORMAP[2]])
if SAVE_PLOTS_AS_PNG:
    plt.savefig("fig3-x-losses.png")
if SAVE_PLOTS_AS_SVG:
    plt.savefig("fig3-x-losses.svg", bbox_inches='tight')
../_images/f9301d5edd91e6c8a057cbf12c22ccbaef790420d89d45fb36daa1ca5e1cea88.png