forked from DmitryUlyanov/deep-image-prior
-
Notifications
You must be signed in to change notification settings - Fork 1
/
plot_dip.py
323 lines (278 loc) · 15.2 KB
/
plot_dip.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
import read_data as rd
import matplotlib.pyplot as plt
import numpy as np
import os
import argparse
import dip_func as df
from mdt_calculations.plotting import geo_plot_w_subplots as gp
import read_data as rd
from PIL import Image
from skimage.transform import resize
from scipy.interpolate import make_interp_spline
from matplotlib.patches import Circle
from skimage.metrics import peak_signal_noise_ratio
from skimage.metrics import structural_similarity as ssim
def plot_from_coord(arr, coord, window=20, vmin=0, vmax=1.4):
fig, ax = plt.subplots()
ax.imshow(arr[coord[0]-window:coord[0]+window, coord[1]-window:coord[1]+window], vmin=vmin, vmax=vmax, cmap='turbo')
circ = Circle(coord, 50)
ax.add_patch(circ)
plt.show()
def compute_std_from_coords(arrs, coords, window=3):
std_coords = []
for coord in coords:
coords_across_arrs = []
for arr in arrs:
coords_across_arrs.append(arr[coord[0]-window:coord[0]+window, coord[1]-window:coord[1]+window])
std_coords.append(np.std(coords_across_arrs))
return std_coords
def compute_std_across_arrs(arrs):
arr_stack = np.stack(arrs)
return np.std(arr_stack, axis=0)
def compute_psnr_against_reference_coords(arrs, reference, coords, window=5):
errors = []
for arr in arrs:
ref_windows = []
pred_windows = []
for coord in coords:
ref_windows.append(reference[coord[0]:coord[0]+window, coord[1]:coord[1]+window])
pred_windows.append(arr[coord[0]:coord[0]+window, coord[1]:coord[1]+window])
ref_window_arr = np.stack(ref_windows)
pred_window_arr = np.stack(pred_windows)
error = peak_signal_noise_ratio(ref_window_arr, pred_window_arr, data_range=10)
errors.append((error))
return np.array(errors)
def compute_ssim_against_reference_coords(arrs, reference, coords, window=7):
errors = []
for arr in arrs:
ssims = []
for coord in coords:
ref_window = reference[coord[0]:coord[0]+window, coord[1]:coord[1]+window]
pred_window = arr[coord[0]:coord[0]+window, coord[1]:coord[1]+window]
ssim_val = ssim(ref_window, pred_window, data_range=10, win_size=min(7, window))
ssims.append(ssim_val)
error = np.mean(ssims)
errors.append(error)
return np.array(errors)
def compute_errors_against_reference(arrs, reference):
errors = []
for arr in arrs:
error = np.mean((arr - reference) ** 2)
errors.append(error)
return np.array(errors)
parser = argparse.ArgumentParser()
parser.add_argument("-i", "--iterations", type=int, default=2000, help="number of iterations")
parser.add_argument("-k", "--kernel_size", type=int, default=0, help="kernel size for mean reg")
parser.add_argument("-r", "--run_number", type=int, default=5, help="chosen run number")
parser.add_argument("-l1", "--lambda_mean_reg", type=float, default=0.0, help="strength of mean reg")
parser.add_argument("--model_key", type=str, default="DIP", help="name of model")
parser.add_argument("--region", type=str, default="global", help="region size on which model was trained")
parser.add_argument("--mdt_filename", type=str, default="dtu18_gtim5_do0280_rr0004.dat", help="name of mdt to denoise")
parser.add_argument("--from_checkpoint", dest="from_checkpoint", action="store_true", help="train from last save")
parser.add_argument("-re", "--region_extent", type=str, default="global", help="region extent of plot")
parser.add_argument("-p", "--product", type=str, default="mdt", help="product type of plot: mdt or cs")
parser.add_argument("--log", dest="log", action="store_true", help="log scale")
args = parser.parse_args()
num_iter = args.iterations
run_number = args.run_number
k = args.kernel_size
region = args.region
model_key = args.model_key
lambda_mean_reg = args.lambda_mean_reg
region_extent = args.region_extent
product = args.product
log = args.log
filepath = os.path.join('data', args.mdt_filename)
# modify me when/if new params are added
param_string = f"region={region}-k={k}-l1={lambda_mean_reg}"
# save format = modelname / region&kernel / mdtname
save_dir = os.path.join('D:/deep-image-prior/results/', model_key, region, param_string, args.mdt_filename[:-4])
save_dir = os.path.join(save_dir, str(run_number))
central_lon = 0
bds = {'mdt': [-1.4, 1.4], 'cs': [0, 1.4]}
vmin, vmax = bds[product]
num_iters = np.arange(100, 4400, 100)
list_dict = []
arr_of_dips = []
for num_iter in num_iters:
if product == 'cs':
arr_of_dips.append(np.flipud(np.rot90(rd.read_surface(f'{num_iter}_cs_band20.dat',
os.path.join(save_dir, 'currents')))))
elif product == 'mdt':
arr_of_dips.append(np.flipud(np.rot90(rd.read_surface(f'{num_iter}.dat', os.path.join(save_dir, 'dats')))))
dict_item = {'extent': f'{region_extent}', 'product': f'{product}', 'log': log}
list_dict.append(dict_item)
iteration_figsize = (17, 10)
standard_figsize = (16, 16)
xlabels = ['(a)', '(b)', '(c)', '(d)', '(e)', '(f)', '(g)', '(h)']
xlabels = [str(item) for item in num_iters]
prod_dict = {'cs': ['cs', 'currents', '_cs_band10'], 'mdt': ['mdt', 'mdts', '']}
if product == 'cs':
cls_name = 'cls18'
else:
cls_name = 'cls18_rr0004'
cls18 = np.rot90(rd.read_surface(f'{cls_name}{prod_dict[product][2]}.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/a_mdt_data/computations/{prod_dict[product][1]}'))
nemo = np.rot90(rd.read_surface(f'orca0083_mdt_12th{prod_dict[product][2]}.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/a_mdt_data/computations/{prod_dict[product][1]}'))
plot_reference = False
if plot_reference:
gp.plot(cls18, product=f'{prod_dict[product][0]}', vmin=vmin, vmax=vmax, central_lon=180)
gp.plot(nemo, product=f'{prod_dict[product][0]}', vmin=vmin, vmax=vmax, central_lon=180)
cls18_res4 = resize(cls18, (720, 1440), order=2)
nemo_res4 = resize(nemo, (720, 1440), order=2)
plot_original = False
orig_mdt = np.rot90(rd.read_surface(f'dtu18_gtim5_do0280_rr0004{prod_dict[product][2]}.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/a_mdt_data/computations/{prod_dict[product][1]}'))
if plot_original:
gp.plot(np.rot90(orig_mdt), product=f'{prod_dict[product][0]}', vmin=0, vmax=2, central_lon=180)
plot_iterations_psnr = False
if plot_iterations_psnr:
coords = [(229, 1131), (248, 1120), (263, 1090), (250, 1167), (576, 1260), (40, 260), (600, 540)]
cls_errors = compute_psnr_against_reference_coords(arr_of_dips, cls18_res4, coords, window=5)
nemo_errors = compute_psnr_against_reference_coords(arr_of_dips, nemo_res4, coords, window=5)
avg_errors = (cls_errors + nemo_errors)/2
cls_glb_errors = compute_errors_against_reference(arr_of_dips, cls18_res4)
nemo_glb_errors = compute_errors_against_reference(arr_of_dips, cls18_res4)
num_iters_spline = np.linspace(num_iters.min(), num_iters.max(), 150)
cls_errors_spline = make_interp_spline(num_iters, cls_errors)
cls_errors_spline = cls_errors_spline(num_iters_spline)
nemo_errors_spline = make_interp_spline(num_iters, nemo_errors)
nemo_errors_spline = nemo_errors_spline(num_iters_spline)
avg_errors = make_interp_spline(num_iters, avg_errors)
avg_errors = avg_errors(num_iters_spline)
fig, ax = plt.subplots(1, 1)
ax.plot(num_iters_spline, cls_errors_spline, linestyle='--', label='CNES-CLS18 Error', linewidth=1)
ax.plot(num_iters_spline, nemo_errors_spline, linestyle='--', label='NEMO Error', linewidth=1)
ax.plot(num_iters_spline, avg_errors, label='Average Error', linewidth=1)
ax.axvline(x=800, color='0.8', linestyle='dashed', linewidth=1.5)
ax.axvline(x=3200, color='0.8', linestyle='dashed', linewidth=1.5)
ax.set_xlabel('Iterations')
ax.set_ylabel('PSNR')
ax.set_xlim([0, 4400])
ax.grid()
ax.legend()
plt.show()
plot_iterations_ssim = False
if plot_iterations_ssim:
coords = [(229, 1131), (248, 1120), (263, 1090), (250, 1167), (576, 1260), (40, 260), (600, 540)]
cls_errors = compute_ssim_against_reference_coords(arr_of_dips, cls18_res4, coords, window=3)
nemo_errors = compute_ssim_against_reference_coords(arr_of_dips, nemo_res4, coords, window=3)
avg_errors = (cls_errors + nemo_errors)/2
cls_glb_errors = compute_errors_against_reference(arr_of_dips, cls18_res4)
nemo_glb_errors = compute_errors_against_reference(arr_of_dips, cls18_res4)
num_iters_spline = np.linspace(num_iters.min(), num_iters.max(), 150)
cls_errors_spline = make_interp_spline(num_iters, cls_errors)
cls_errors_spline = cls_errors_spline(num_iters_spline)
nemo_errors_spline = make_interp_spline(num_iters, nemo_errors)
nemo_errors_spline = nemo_errors_spline(num_iters_spline)
avg_errors = make_interp_spline(num_iters, avg_errors)
avg_errors = avg_errors(num_iters_spline)
fig, ax = plt.subplots(1, 1)
ax.plot(num_iters_spline, cls_errors_spline, linestyle='--', label='CNES-CLS18 Error', linewidth=1)
ax.plot(num_iters_spline, nemo_errors_spline, linestyle='--', label='NEMO Error', linewidth=1)
ax.plot(num_iters_spline, avg_errors, label='Average Error', linewidth=1)
ax.axvline(x=800, color='0.8', linestyle='dashed', linewidth=1.5)
ax.axvline(x=3200, color='0.8', linestyle='dashed', linewidth=1.5)
ax.set_xlabel('Iterations')
ax.set_ylabel('SSIM')
ax.set_xlim([0, 4400])
ax.grid()
ax.legend()
plt.show()
coords_for_stds = [(238, 1120), (228, 540), (511, 99), (531, 1205), (263, 1090), (576, 1260)]
stds_coords = compute_std_from_coords(arr_of_dips, coords_for_stds)
# Plot to see where coordinate is
# for coord in coords_for_stds:
# plot_from_coord(arr_of_dips[17], coord)
plot_std = True
if plot_std:
# get rid of cbar title for this
std_arr_stack = compute_std_across_arrs(arr_of_dips[7:31])
list_dicts_stds = [{'extent': 'gs_big', 'product': 'std'},
{'extent': 'ku_big', 'product': 'std'},
{'extent': 'ag', 'product': 'std'},
{'extent': 'bm3', 'product': 'std'}]
stds_labels = ['(a)', '(b)', '(c)', '(d)']
gp.geo_subplots([std_arr_stack, std_arr_stack, std_arr_stack, std_arr_stack],
list_dicts_stds, 2, 2, vmin=0, vmax=0.4,
figsize=iteration_figsize, xlabels=stds_labels, ticksize=10)
# gp.plot(std_arr_stack, product=f'std', vmin=0, vmax=0.4, central_lon=180, extent='gs_big')
# gp.plot(std_arr_stack, product=f'std', vmin=0, vmax=0.4, central_lon=180, extent='ku_big')
# gp.plot(std_arr_stack, product=f'std', vmin=0, vmax=0.4, central_lon=180, extent='bm')
# gp.plot(std_arr_stack, product=f'std', vmin=0, vmax=0.4, central_lon=180, extent='ag')
# Residual plot
# gp.plot((arr_of_dips[21]-arr_of_dips[17]), product=f'{prod_dict[product][0]}', vmin=0, vmax=0.5, central_lon=180, cmap='viridis')
# Average over less iterations
# gp.plot(np.mean(arr_of_dips[16:22], axis=0), product=f'{prod_dict[product][0]}', vmin=0, vmax=2, central_lon=180)
num_iters_window_plot = np.arange(200, 4400, 400)
num_iters_window_plot = np.array([100, 200, 400, 800, 1200, 1600, 2000, 2400, 2800, 3200, 3600, 4000])
list_dict_window_plot = []
arr_dips_window = []
xlabels_window = [str(item) for item in num_iters_window_plot]
for num_iter in num_iters_window_plot:
if product == 'cs':
arr_dips_window.append(np.flipud(np.rot90(rd.read_surface(f'{num_iter}_cs_band20.dat',
os.path.join(save_dir, 'currents')))))
elif product == 'mdt':
arr_dips_window.append(np.flipud(np.rot90(rd.read_surface(f'{num_iter}.dat', os.path.join(save_dir, 'dats')))))
dict_item = {'extent': f'{region_extent}', 'product': f'{product}', 'log': log}
list_dict_window_plot.append(dict_item)
plot_snapshot_over_iterations = False
if plot_snapshot_over_iterations:
gp.geo_subplots(arr_dips_window, list_dict_window_plot, 3, 4, vmin=vmin, vmax=vmax, figsize=iteration_figsize, cbarlabelsize=14,
ticksize=12, xlabels=xlabels_window)
plot_snapshot = True
list_dicts_snapshot = [{'extent': 'gs', 'product': 'cs'},
{'extent': 'ku', 'product': 'cs'},
{'extent': 'ag', 'product': 'cs'},
{'extent': 'bm', 'product': 'cs'}]
snapshot_labels = ['(a)', '(b)', '(c)', '(d)']
if plot_snapshot:
coords = [(103, 32), (-45, 32), (-155, -35), (121, -43), (-50, -40), (-108, -36), (116, 28), (-40, 25)]
gp.plot(arr_of_dips[15], product=product, vmin=vmin, vmax=vmax, extent='global', markers=coords)
gp.geo_subplots([arr_of_dips[15], arr_of_dips[15], arr_of_dips[15], arr_of_dips[15]],
list_dicts_snapshot, 2, 2, vmin=vmin, vmax=vmax,
figsize=iteration_figsize, xlabels=snapshot_labels)
# gp.polar_plot(arr_of_dips[15], product=product, vmin=vmin, vmax=vmax)
plot_avg = False
avg_iterations = np.mean(arr_of_dips[14:19], axis=0)
if plot_avg:
gp.plot(avg_iterations, product=product, vmin=vmin, vmax=vmax, extent='global')
gp.polar_plot(avg_iterations, product=product, vmin=vmin, vmax=vmax)
if product == 'cs':
gauss_mdt = np.rot90(rd.read_surface('dtu18_GTIM_R5_do0280_rr0004_100k_cs.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/a_mdt_data/computations/{prod_dict[product][1]}/gauss_filtered_mdts_cs/geodetic'))
else:
gauss_mdt = np.rot90(rd.read_surface('dtu18_GTIM5_R6e_do0280_rr0004_100k.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/a_mdt_data/computations/{prod_dict[product][1]}/gauss_filtered_mdts'))
plot_gaussian = False
if plot_gaussian:
gp.plot(gauss_mdt, product=product, vmin=vmin, vmax=vmax, extent='global')
list_dicts_residual = [{'extent': 'gs', 'product': 'mdt', 'vmin': vmin, 'vmax': vmax},
{'extent': 'gs', 'product': 'mdt', 'vmin': vmin, 'vmax': vmax},
{'extent': 'gs', 'product': 'residual', 'vmin': -0.5, 'vmax': 0.5}]
resid_labels = ['(a)', '(b)', '(c)']
plot_residual = True
if plot_residual:
gp.geo_subplots([orig_mdt, avg_iterations, orig_mdt-avg_iterations],
list_dicts_residual, 1, 3, vmin=None, vmax=None,
figsize=iteration_figsize, xlabels=resid_labels)
plot_dip_gauss_residual = True
if plot_dip_gauss_residual:
gp.geo_subplots([avg_iterations, gauss_mdt, avg_iterations-gauss_mdt],
list_dicts_residual, 1, 3, vmin=None, vmax=None,
figsize=iteration_figsize, xlabels=resid_labels)
# df.compute_currents_from_dat(f'../a_mdt_data/computations', 'dtu18_gtim5_do0280_rr0004')
if product =='cs':
pmf_mdt = np.flipud(np.rot90(rd.read_surface('fmdt_01_L280_55_cs_band20.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/MDT-Calculations/pmf_stuff/io/{prod_dict[product][1]}')))
else:
pmf_mdt = np.flipud(np.rot90(rd.read_surface('fmdt_01_L280_82.dat',
f'C:/Users/oa18724/Documents/Master_PhD_folder/MDT-Calculations/pmf_stuff/io/{prod_dict[product][1]}')))
# pmf_mdt = resize(pmf_mdt, (720, 1440), order=2)
plot_dip_pmf_residual = True
if plot_dip_pmf_residual:
gp.geo_subplots([avg_iterations, pmf_mdt, avg_iterations-pmf_mdt],
list_dicts_residual, 1, 3, vmin=-0.7, vmax=0.7,
figsize=iteration_figsize, xlabels=resid_labels)