from pathlib import Path
from rich import print
from pyinspect._colors import mocassin, orange
print(f'[{mocassin}]Generating [{orange}]figures[/{orange}][{mocassin}]...\n')
metadata = load_yaml('figures/composed_figures.yml')
figs_folder = Path('figures')
for figname, figs in metadata.items():
if len(figs) < 2:
ncols, nros = 1, 1
elif len(figs) < 3:
ncols, nrows = 2, 1
elif len(figs) < 5:
ncols, nrows = 2, 2
elif len(figs) < 7:
ncols, nrows = 3, 2
f, axarr = plt.subplots(ncols=ncols, nrows=nrows, figsize=(18, 13))
f.suptitle(figname)
for ax in axarr.flatten():
ax.axis('off')
for ax, (title, fig) in zip(axarr.flatten(), figs):
ax.imshow(mpimg.imread(figs_folder / fig))
ax.set(title=title)
f.tight_layout()
save_figure(f, figs_folder / figname, verbose=False)
xticks=[],
xlim=[x.min(), x.max()],
)
clean_axes(f)
axes, titles = tag_axes, ('stim', 'start', 'run', 'shelter', 'stop')
for ax, ttl in zip(tag_axes, titles):
ax.set(title=ttl)
ax.set(xticks=[0, 30, 60],
xticklabels=[-1, 0, 1],
xlabel='s',
yticks=[],
ylim=axarr[1, 0].get_ylim())
ax.spines['left'].set_visible(False)
ax.axvline(30, color=[.2, .2, .2], lw=3)
save_figure(
f,
fld /
f'{sess["mouse"]}_{sess["date"]}__{roi}_active_{ACTIVE["active"][-1]}',
verbose=False)
# plt.close(f)
break
break
pd.DataFrame(ACTIVE).to_hdf('ACTIVE_ROIS.h5', key='hdf')
# %%
# %%
# ax.plot(_time, signals[n][stim-frames_pre:at_shelter], c=signal_color, alpha=.3)
# Clean up figure
ttls = ['Shelter distance', 'speed', 'angular velocity']
for n, ax in enumerate(axarr):
ax.axvline(0, lw=2, ls='--', color=[.4, .4, .4])
if n < len(ttls):
ttl = ttls[n]
else:
ttl = 'signal'
ax.set(ylabel=ttl)
axarr[-1].set(xlabel='frames')
clean_axes(f)
save_figure(f, str(save_fld / f'{mouse} {sess} behav traces'))
# break
# %%
"""
Plot each trials baseline as a scatter plot + scatter of PCA
"""
def fit_lin_regrs(R, X, S):
"""
predicts a ROI's mean baseline activity at each trial (R)
with the mean X position (X) and mean speed (S) at each trial
with a linear regression + OLS
"""
overal_accuracies[name] = accuracy
# Clean axes
axarr[s].legend()
s += 1
# break
#
axarr[-1].axis('off')
clean_axes(f)
set_figure_subplots_aspect(hspace=.6, top=.9, wspace=.25)
save_figure(
f,
str(save_fld /
f'Shelter representation {"stationary" if KEEP_ONLY_STATIONARY else ""} {PREDICT}'
))
# %%
# Fit SVM to individual ROIs
f, axarr = plt.subplots(ncols=3, nrows=3, figsize=(16, 16))
axarr = axarr.flatten()
s = 0
f.suptitle(f'individual ROI SVM accuracy (shelter detection)')
for mouse, sess, name in tqdm(mouse_sessions):
if pcas[name] is None: continue
# Prep data
N = 500
scatter_trial_baseline(axarr[n, 1], WHOLE_TRACE_DFF[:n_frames_pre], count, len(tags))
# Plot signals
axarr[n, 0].plot(RAW, color='skyblue', lw=2, alpha=.6)
axarr[n, 1].plot(WHOLE_TRACE_DFF, color='salmon', lw=2, alpha=.6)
axarr[n, 2].plot(DORIC_CHUNK_DFF, color='seagreen', lw=2, alpha=.6)
axarr[n, 3].plot(BASELINE_DFF, color='magenta', lw=2, alpha=.6)
# Mark tag onset
for ax in axarr[n, :]:
ax.axvline(n_frames_pre, lw=2, color='r')
# make titles, axis labels etc
ts = [f' RAW Tag: {tag_type} ', ' WHOLE SESS DFF', ' DORIC CHUNK DFF', ' PRE STIM BASELINE DFF']
ttls = [rid + t for t in ts]
ylbs = ['raw', 'dff', 'dff', 'dff']
for a, ax in enumerate(axarr[n, :]):
ax.set(title=ttls[a], ylabel=ylbs[a])
# Set X labels
for ax in axarr[-1, :]:
ax.set(xlabel='frames')
# clean up and save
clean_axes(f)
f.tight_layout()
save_figure(f, fld / (sessname))
sig[stim_frame],
color='k',
s=100,
zorder=99)
axarr[n].legend()
axarr[n].set(title=rid, ylabel='raw signal')
# break
# cleanup and save
axarr[-1].set(xlabel='frames')
clean_axes(f)
f.tight_layout()
save_figure(f, fld / (sessname + '_doric_chunks_baselines'))
break
# %%
"""
Plot each ROI's signal with the chunk smoothed by a rolling mean filter
"""
window = 600
for mouse, sess, sessname in track(mouse_sessions,
description='plotting smoothed chunks'):
if not DO['plot_chunks_smoothed']: break
# Get data
tracking, ang_vel, speed, shelter_distance, dffs, signals, nrois, is_rec, roi_ids = \
get_mouse_session_data(mouse, sess, sessions, hanning_window=hanning)
tags = get_tags_by(mouse=mouse,
# fix figure
for n, ax in enumerate(axarr[2:]):
if n >= len(signals):
ax.axis('off')
else:
make_signal_ax(ax, f'ROI {n}', start, end)
axarr[0].set(title='tracking', xlabel='X (px)', ylabel='Y (px)')
axarr[0].legend(handles=legend_elements)
axarr[1].set(title='tracking', xlabel='speed (cm/s)', ylabel='X (px) ')
clean_axes(f)
set_figure_subplots_aspect(right=.85, top=.9, wspace=.4)
save_figure(
f,
str(fld /
f'{sessname}_tag_{tag_type[-1]}_{"".join(event_type)}_mean_{HIGHLIGHT_MEAN}'
),
svg=False)
# break
with open(str(fld / '00_info.txt'), 'w') as out:
out.write('INFO')
out.write(f'''IMAGES GENERATED WITH PARAMS:
fps = {fps }
n_sec_pre = {n_sec_pre}
n_sec_post = {n_sec_post}
tag_type = {tag_type}
event_type = {event_type}
NORMALIZE = {NORMALIZE}
FILTER = {FILTER}
keys = []
plot_model(axes['escape_peak_speed'][4],
ols_behav_traces_escape_peak_speed, *args)
plot_model(axes['escape_onset'][4],
ols_behav_traces_escape_onset, *args)
plot_model(axes['stim_onset'][4], ols_behav_traces_escape_stim,
*args)
plot_model(axes['shelter_arrival'][4],
ols_behav_traces_shelter_arrival, *args)
plot_model(axes['homing'][4], ols_behav_traces_homing_onset,
*args)
plot_model(axes['homing_peak_speed'][4],
ols_behav_traces_hominst_peak_speed, *args)
# Refine figure
f.suptitle(
f"{mouse} - {sess} - {roi_ids[n]} - {tot_trials} trials")
for ax in f.axes:
adjust_ticks(ax, frames_pre, frames_post, fps, every=1)
ax.set(xlim=[xmin, xmax])
# Save and close
save_figure(f, f_path, verbose=True)
if not DEBUG: plt.close(f)
if DEBUG: break
if DEBUG: break
if DEBUG: break
plt.show()
xlim=[-600, 600])
meanspeedsax.set(title='Bout mean ang vel and speed',
xlabel='speed (px/s)',
ylabel='angular velocity (deg/s)',
ylim=[-100, 100],
xlim=[-600, 600])
meanspeedsax.axhline(0, color='k', alpha=.5, lw=4, ls='--')
meanspeedsax.axvline(0, color='k', alpha=.5, lw=4, ls='--')
meanspeedsax.legend()
framespeedsax.axhline(0, color='k', alpha=.5, lw=4, ls='--')
framespeedsax.axvline(0, color='k', alpha=.5, lw=4, ls='--')
framespeedsax.legend()
speedsax.set(title='Avg bout speed distribution',
xlabel='speed',
ylabel='density')
speedsax.legend()
# Clean and save figure
set_figure_subplots_aspect(wspace=.4, hspace=.4, top=.9, bottom=.15)
clean_axes(f)
f.suptitle(f'{mouse} summary.', fontsize=22)
save_figure(
f,
os.path.join(output_fld,
f'{mouse}_highspeed_{high_speed_bouts}_bouts_summary'))
# break
# %%
ax.plot(X, speed, lw=3, color=salmon_dark, label='seed', alpha=.5)
ax.plot(X,
smoothed - smoothed[rec_off],
lw=3,
color=light_blue_dark,
label='DFF',
alpha=.5)
ax.legend()
ax.set(title=rid, xlabel='frames', ylabel='norm.speed and dff')
clean_axes(f)
f.tight_layout()
save_figure(f, sfld / f'{rid}_dff_vs_speed')
# break
# break
# %%
'''
Make histogram of pearson correlations between filtered variables
'''
speed_th = .2
window = 30
pvals = [[], []]
speed_coeffs, dist_coffs, speed_dist_coeffs, pc_speed_coeffs = [[], []
], [[], []
], [[], []
axarr[0].axhline(liftoff_th, lw=6, color='k', alpha=.2, ls='--')
axarr[0].axhline(0, lw=2, color='k', zorder=-1)
axarr[0].legend()
axarr[1].legend()
axarr[0].set(ylabel='$W_{pc}$')
axarr[1].set(ylabel='$\\frac{d W_{pc}}{dt}$',
**xkwargs,
xlabel='time (ms)',
xlim=[-2, 2])
# save figure
clean_axes(fig)
save_figure(fig,
str(save_fld / f"trial_plot_{trial['name']}"),
verbose=False)
if i > 0:
plt.close(fig)
# if i == 2:
# break
save_figure(main_fig, str(save_fld / f"all_trials"), verbose=False)
if len(step_start_frames) == len(data):
data['step_start_frames'] = step_start_frames
# %%
from sklearn.decomposition import PCA
from einops import rearrange
clean_axes(f)
for row in range(3):
maxy = np.max([ax.get_ylim()[1] for ax in axarr[row, :]])
miny = np.min([ax.get_ylim()[0] for ax in axarr[row, :]])
for n, ax in enumerate(axarr[row, :]):
ax.set(ylim=[miny, maxy])
if n > 0:
ax.spines['left'].set_visible(False)
ax.set(yticks=[])
if row < 2:
ax.spines['bottom'].set_visible(False)
save_figure(f,
fld / f'{sess["mouse"]}_{sess["date"]}__{roi}',
verbose=False)
plt.close(f)
del rois
# break
# break
# %%
# ''' Plot the mean trace aligned to each tag for each ROI and the first PCA component'''
# f, axarr = plt.subplots(ncols=5, nrows=2, figsize=(16, 9))
# for sessn, sess in enumerate(Sessions.fetch(as_dict=True)):
# print(sess['mouse'], sess['date'])
# data, rois = get_session_data(sess['mouse'], sess['date'], roi_data_type='raw')
# if data is None: continue
points = cont['coordinates']
axarr[0].plot(*points.T, color=color, lw=2)
axarr[1].plot(*points.T, color=color, lw=2, alpha=alpha)
axarr[1].text(cont['CoM'][1],
cont['CoM'][0],
str(n),
color=txtcolor,
zorder=100)
axarr[2].imshow(1 - np.sum(new_masks, 2), cmap="gray_r")
axarr[0].set(title='A', xticks=[], yticks=[])
axarr[1].set(title=f'Masks - threhsold {mask_th}', xticks=[], yticks=[])
save_figure(f, str(fld / 'contours'))
# Save new masks and contours to file
np.save(str(fld / 'masks.npy'), new_masks)
pd.DataFrame(contours).to_hdf(str(fld / 'contours.h5'), key='hdf')
# %%
# ---------------------------------------------------------------------------- #
# EXTRACT TRACES #
# ---------------------------------------------------------------------------- #
n_frames = video.shape[0]
traces = np.zeros((n_frames, n_components))
for compn in track(range(n_components)):
traces[:, compn] = get_component_trace_from_video(compn, masks, n_frames,