EquiFusion / src /utils /plot_parallel.py
ceccurr's picture
EquiFusion demo
e6d357a verified
Raw History Blame Contribute Delete
7.09 kB
import matplotlib.pyplot as plt
from matplotlib import animation
import numpy as np
from .plot import create_pose
RESOPTIONS = {}
RESOPTIONS["linewidth"] = 2
RESOPTIONS["figsize"] = 6
def set_axes_equal(ax):
"""
Make axes of 3D plot have equal scale so that spheres appear as spheres,
cubes as cubes, etc.
Input
ax: a matplotlib axis, e.g., as output from plt.gca().
"""
ax.set_xlim3d([-1, 1])
ax.set_ylim3d([-1, 1])
ax.set_zlim3d([-1, 1])
def plot_skeleton(vals, limbseq, ax, update=False, plots=None, color="g"):
# Start and endpoints of our representation
I = np.array([touple[0] for touple in limbseq])
J = np.array([touple[1] for touple in limbseq])
if plots is None:
plots = []
for i in np.arange(len(I)):
x = np.array([vals[I[i], 0], vals[J[i], 0]])
y = np.array([vals[I[i], 1], vals[J[i], 1]])
z = np.array([vals[I[i], 2], vals[J[i], 2]])
if not update:
plots.append(
ax.plot(x, y, z, lw=RESOPTIONS["linewidth"], color=color)[0]
) # , label=['GT' if not pred else 'Pred']) #lw=2, linestyle='--',
elif update:
plots[i].set_xdata(x)
plots[i].set_ydata(y)
plots[i].set_3d_properties(z)
plots[i].set_color(color)
return plots
def get_drawing_funct(
kpts3d_obs, kpts3d_gt, kpts3d_all, plots, axes, skeleton, keep_gt=False
):
kwargs = {"limbseq": skeleton.limbseq, "left_right_limb": skeleton.left_right_limb}
n_limbs = len(kwargs["limbseq"])
def iterate_over_2daxes(axes):
for axarr in axes:
for ax in axarr:
yield ax
def plot_multiple_pred(ax, ax_plot, ax_kpts_list, n, keep_gt=False):
assert not keep_gt
colors = ["g", "b"]
# change color of first limb segments
create_pose(
ax,
ax_plot,
vals=ax_kpts_list[0][n],
pred=False,
num_multiple_preds=1,
update=True,
random_color=False,
color="r",
**kwargs,
)
# plot other predictions
for i, pred in enumerate(ax_kpts_list[1:]):
if len(ax_plot) < ax_kpts_list[0].shape[-2] * (i + 1):
create_pose(
ax,
ax_plot,
vals=pred[n],
pred=True,
num_multiple_preds=i + 1,
update=False,
random_color=False,
color=colors[i],
**kwargs,
)
else:
create_pose(
ax,
ax_plot,
vals=pred[n],
pred=True,
num_multiple_preds=i + 1,
update=True,
random_color=False,
color=colors[i],
**kwargs,
)
def plot_gt_and_pred(ax, ax_plots, ax_kpts_gt, ax_kpts, n):
# first get target gt
create_pose(
ax,
ax_plots,
vals=ax_kpts_gt[n],
pred=False,
num_multiple_preds=1,
update=True,
random_color=False,
**kwargs,
)
# and closestGT overlapped
if len(ax_plots) < n_limbs + 2:
create_pose(
ax,
ax_plots,
ax_kpts[0],
pred=True,
update=False,
num_multiple_preds=1,
**kwargs,
)
else:
create_pose(
ax,
ax_plots,
vals=ax_kpts[n],
pred=True,
num_multiple_preds=1,
update=True,
random_color=False,
**kwargs,
)
def draw_pred_ontopof_obs(ax, ax_plots, ax_kpts, n):
create_pose(
ax,
ax_plots,
vals=ax_kpts[n],
pred=False,
num_multiple_preds=1,
update=True,
random_color=False,
color="r",
**kwargs,
)
def drawframe(n):
for ax in iterate_over_2daxes(axes):
ax.set_title(
ax.get_title(loc="center").split("\n")[0]
+ f"\nframe {n+1}/{len(kpts3d_obs) + len(kpts3d_gt)}"
)
if n < len(kpts3d_obs):
# draw obervation
for i, ax in enumerate(iterate_over_2daxes(axes)):
ax_plots = plots[i]
create_pose(
ax,
ax_plots,
vals=kpts3d_obs[n],
pred=False,
num_multiple_preds=1,
update=True,
random_color=False,
color=None,
**kwargs,
)
else:
# draw prediction
n -= len(kpts3d_obs)
for i, (ax_plots, ax, vals) in enumerate(
zip(plots, iterate_over_2daxes(axes), kpts3d_all)
):
if i == 0 or keep_gt:
plot_gt_and_pred(ax, ax_plots, kpts3d_gt, vals, n)
else:
draw_pred_ontopof_obs(ax, ax_plots, vals, n)
return plots
return drawframe
def create_plot_canvas(
kpts3d_obs, plot_titles, figsize, skeleton, n_rows=None, n_cols=None
):
kwargs = {"limbseq": skeleton.limbseq, "left_right_limb": skeleton.left_right_limb}
fig = plt.figure(
figsize=(figsize, figsize)
) # change here for speed & lightweight, 20
label = "Multiple Predictions"
nrows = (
int(len(plot_titles) // 3) + int(len(plot_titles) % 3 != 0)
if n_rows is None
else n_rows
)
ncols = 3 if n_cols is None else n_cols
fig, axes = plt.subplots(
nrows=nrows, ncols=ncols, subplot_kw=dict(projection="3d"), figsize=(16, 16)
)
title = f"Prediction {label}"
fig.suptitle(title, fontsize=16)
if nrows == 1:
axes = np.expand_dims(axes, axis=0)
for axarr in axes:
for ax in axarr:
ax.set_box_aspect([1, 1, 1])
set_axes_equal(ax)
ax.set_xlabel("x")
ax.set_ylabel("z")
ax.set_zlabel("y")
if nrows == 1:
break
plots = []
for i, axarr in enumerate(axes):
for j, ax in enumerate(axarr):
plots.append(
create_pose(
ax,
[],
vals=kpts3d_obs[0],
pred=False,
num_multiple_preds=1,
update=False,
**kwargs,
)
)
idx = i * ncols + j
ax.set_title(plot_titles[idx] if idx < len(plot_titles) else "")
ax.view_init(elev=15.0, azim=70)
if nrows == 1:
break
return fig, axes, plots