def experiment(*, max_count, drift_velocity, drift_angle, **kwargs):
sv = StrandVideo(angle_velocity=0)
N = len(sv.frames)
corr_sum = correlate_and_sum(sv.frames)
drift_new1, d = guizar_multiframe(corr_sum, start=1, end=N)
drift_old = np.mean(d[9:29], axis=0)
drift_new2 = guizar_upsample(corr_sum)[0]
# import ipdb
# ipdb.set_trace()
return {
'error_old': np.linalg.norm(sv.true_drift - drift_old),
'error_new1': np.linalg.norm(sv.true_drift - drift_new1),
'error_new2': np.linalg.norm(sv.true_drift - drift_new2),
'drift_old': drift_old,
'drift_new1': drift_new1,
'drift_new2': drift_new2,
'errorv_old': sv.true_drift - drift_old,
'errorv_new1': sv.true_drift - drift_new1,
'errorv_new2': sv.true_drift - drift_new2,
'd': d,
'd_err': np.linalg.norm(sv.true_drift - d, axis=1),
'true_drift': sv.true_drift
}
def experiment(*, max_count, background_noise, drift_velocity):
# noise
noise_model = get_visors_noise(background=background_noise)
sv = StrandVideo(
scene=scene,
max_count=max_count,
noise_model=noise_model,
drift_velocity=drift_velocity * 1e-3,
)
noise = sv.frames - sv.frames_clean
# register
corr_sum = correlate_and_sum(sv.frames)
drift, _ = guizar_multiframe(corr_sum)
# reconstruct
recon = shift_and_sum(sv.frames, drift, mode='crop')
noise_recon = shift_and_sum(noise, drift, mode='crop')
result = resize(recon, (300, 300))
snr_db = 10 * np.log10(
np.sum((recon - noise_recon)**2) / np.sum(noise_recon**2))
return {
'result': result,
'est_drift': drift,
'true_drift': sv.true_drift,
'snr_db': snr_db,
}
def experiment(*, max_count, background_noise, drift_velocity, frame_rate):
# noise
noise_model = get_visors_noise(background=background_noise)
sv = StrandVideo(
ccd_size=size,
start=((1400, 1300)),
scene=scene,
max_count=max_count,
noise_model=noise_model,
drift_velocity=drift_velocity * 1e-3,
resolution_ratio=resolution_ratio,
fov_ratio=fov_ratio,
frame_rate=frame_rate
)
# register
corr_sum = correlate_and_sum(sv.frames)
drift, _ = guizar_multiframe(corr_sum)
# drift_u, _ = guizar_upsample(corr_sum)
# reconstruct
coadded = shift_and_sum(sv.frames, drift, mode='crop')
coadded = resize(coadded, (300, 300))
frame = resize(sv.frames[0], (300, 300))
return {
'coadded': coadded,
'frame': frame,
'est_drift': drift,
# 'est_drift_u': drift_u,
'true_drift': sv.true_drift,
# 'd': d
}
def experiment(*, drift_velocity, drift_angle, max_count):
sv = StrandVideo(
drift_velocity=drift_velocity,
max_count=max_count,
drift_angle=drift_angle
)
corr_sum = correlate_and_sum3(sv.frames)
gm, d = guizar_multiframe2(corr_sum, start=1, end=39)
return {
'component_errors': np.linalg.norm(sv.true_drift - d, axis=1)
}
def experiment(*, max_count, drift_velocity, drift_angle, **kwargs):
sv = StrandVideo(
max_count=max_count,
drift_velocity=drift_velocity,
drift_angle=drift_angle
)
corr_sum = correlate_and_sum3(sv.frames)
guizar_drift, d = guizar_multiframe2(corr_sum, start=10, end=30)
ulas_drift1, ulas_drift2 = ulas_multiframe3(corr_sum)
return {
'guizar_error': np.linalg.norm(sv.true_drift - guizar_drift),
'guizar_drift': guizar_drift,
'ulas_error1': np.linalg.norm(sv.true_drift - ulas_drift1),
'ulas_error2': np.linalg.norm(sv.true_drift - ulas_drift2),
'ulas_drift1': ulas_drift1,
'ulas_drift2': ulas_drift2,
'd': np.linalg.norm(sv.true_drift - d, axis=1),
'true_drift': sv.true_drift
}
from skimage.transform import resize
from mas.strand_generator import StrandVideo, get_visors_noise
from mas.tracking import guizar_multiframe, correlate_and_sum, shift_and_sum
from mas.misc import combination_experiment
from html_slider.html_slider import render_pandas
import numpy as np
scene = StrandVideo(noise_model=None).scene
def experiment(*, max_count, background_noise, drift_velocity):
# noise
noise_model = get_visors_noise(background=background_noise)
sv = StrandVideo(
scene=scene,
max_count=max_count,
noise_model=noise_model,
drift_velocity=drift_velocity * 1e-3,
)
noise = sv.frames - sv.frames_clean
# register
corr_sum = correlate_and_sum(sv.frames)
drift, _ = guizar_multiframe(corr_sum)
# reconstruct
recon = shift_and_sum(sv.frames, drift, mode='crop')
noise_recon = shift_and_sum(noise, drift, mode='crop')
result = resize(recon, (300, 300))