mpl_stereo

KO | EN

gitlite — search

#javascript #python #react #ai #golang #machine-learning #deep-learning #linux #nodejs #awesome #cli #kubernetes #mcp #ios

(★ 39)

See your matplotlib plots in 3D by making stereograms, anaglyphs, and wigglegrams.

mpl_stereo 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

![Release](https://img.shields.io/github/v/release/scottshambaugh/mpl_stereo?sort=semver)
[![Downloads](https://static.pepy.tech/badge/mpl_stereo)](https://pepy.tech/project/mpl_stereo)
![Builds](https://github.com/scottshambaugh/mpl_stereo/actions/workflows/builds.yml/badge.svg)
![Tests](https://github.com/scottshambaugh/mpl_stereo/actions/workflows/tests.yml/badge.svg)
[![codecov](https://codecov.io/gh/scottshambaugh/mpl_stereo/graph/badge.svg?token=V2ZSLFUK03)](https://codecov.io/gh/scottshambaugh/mpl_stereo)
![PyPI - Python Version](https://img.shields.io/pypi/pyversions/mpl_stereo)

Matplotlib add-on to make [stereograms](https://en.wikipedia.org/wiki/Stereoscopy), [anaglyphs](https://en.wikipedia.org/wiki/Anaglyph_3D), and [wigglegrams](https://en.wikipedia.org/wiki/Wiggle_stereoscopy).

Stereographic images can significantly enhance the interpretability of 3D data by leveraging human binocular vision. Instead of looking at a flat projection on a page, stereograms give us "3D glasses" for 2D data with just our eyes.

It takes some practice to be able to view the stereoscopic effect for the first time, but the effort is well worth it!

## Basic Usage
**Python Installation**
```
pip install mpl_stereo
```
**Setup**
```python
from mpl_stereo import AxesStereo2D, AxesStereo3D, AxesAnaglyph
from mpl_stereo.example_data import trefoil
x, y, z = trefoil()  # trefoil knot
```

### 2D Stereogram plots
Currently, only a subset of matplotlib's 2D plotting methods are officially supported: `plot`, `step`, `scatter`, `stem`, `bar`, `fill`, `fill_between`, `errorbar`, `text`, `semilogx`, `semilogy`, and `loglog`.

`fill_between` takes the depth as a `z` keyword (`fill_between(x, y1, y2, z=z)`) since its positional arguments are the two fill boundaries. For the log methods, the `z` depth is always treated as linear. This list is also available at runtime as `axstereo.known_methods`.
```python
axstereo = AxesStereo2D()
axstereo.plot(x, y, z, c='k', alpha=0.2)
axstereo.scatter(x, y, z, c=z, cmap='viridis', s=10)
```

<img width="500" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_2d.png">

When you are viewing the stereogram properly, you should see the knot weave in and out of the page!

*Warning*: Please note that for 2D plots the stereoscopic effect requires shifting data, so the data will *not* necessarily line up with the x-axis labels! Right now this is controlled with the `eye_balance` parameter. Calling `AxesStereo2D(eye_balance=-1)` (the default) will ensure that the left x-axis data is not shifted, whereas `AxesStereo2D(eye_balance=1)` will lock down the right x-axis data. The tick labels for x-axes where the data is not aligned will have transparency applied. So in the plot above, the right side labels being lighter gray indicates that you should not trust that x-axis data to be positioned correctly, but the left subplot with its black labeling is accurate.

### 3D Stereogram plots
The stereoscopic effect in 3D is made just by rotating the plot view, so all of matplotlib's 3D plot types are supported, and there are no concerns about data not lining up with the axis labels.
```python
axstereo = AxesStereo3D()
axstereo.plot(x, y, z, c='k', alpha=0.2)
axstereo.scatter(x, y, z, c=z, cmap='viridis', s=10)
```

<img width="500" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_3d.png">

### Red-Cyan Anaglyphs
Plots can also be made into anaglyphs, which are stereograms that can be viewed with red-cyan 3D glasses. While this allows for seeing the stereoscopic effect without training your eyes, it also means that the data cannot be otherwise colored. Use `AxesAnaglyph` for 2D plots and `AxesAnaglyph3D` for 3D plots.

The same warning as for the 2D stereo plots about shifting data applies here as well. If `eye_balance` is -1 or +1 such that the data for one of the colors is not shifted, then that color will be applied to the x-axis tick labels to show that they are accurate.
```python
axstereo = AxesAnaglyph()
axstereo.plot(x, y, z)
axstereo.scatter(x, y, z, s=10)
```

<img width="250" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_anaglyph.png">

```python
axstereo = AxesAnaglyph3D()
axstereo.plot(x, y, z)
axstereo.scatter(x, y, z, s=10)
```

<img width="250" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_anaglyph_3d.png">

### Existing Stereo Images
Existing stereo image data can easily be plotted up side-by-side, or combined into an anaglyph. Here we plot up two 1-D grayscale images of the sun taken in July 2023 by the NASA STEREO-A and SDO spacecraft. This example was adapted from [the SunPy documentation](https://docs.sunpy.org/en/stable/generated/gallery/showcase/stereoscopic_3d.html#sphx-glr-generated-gallery-showcase-stereoscopic-3d-py).

Every stereogram is written to file with its `save()` method. For images it's often nice to save just the picture without any surrounding axis decorations, which is what `save(..., plot_area=True)` does.
```python
from mpl_stereo.example_data import sun_left_right
sun_left_data, sun_right_data = sun_left_right

axstereo = AxesStereo2D()
axstereo.imshow_stereo([sun_left_data, sun_right_data], cmap='gray')  # try other colormaps!
axstereo.save('sun_2d.png', plot_area=True)  # just the plot area

axstereo = AxesAnaglyph()
axstereo.imshow_stereo([sun_left_data, sun_right_data], cmap='gray')
axstereo.save('sun_anaglyph.png', plot_area=True)
```

<img width="400" height="200" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/sun_2d.png">


<img width="200" height="200" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/sun_anaglyph.png">

Here's another example, showing how this translates to full color data. This example is a pair of photos of [St. Mary's Church](https://commons.wikimedia.org/wiki/File:St_Mary%27s_Church,_Colston_Bassett_3D-35486887876.jpg) in Colston Basset, Britain, taken by David Skinner and shared under a [CC-BY2.0 license](https://creativecommons.org/licenses/by/2.0/deed.en).

Color anaglyph algorithms can be chosen from the methods `'dubois'`, `'photoshop'`, and `'photoshop2'`, as described in the paper [Sanders, William R., and David F. McAllister. "Producing anaglyphs from synthetic images." *Stereoscopic displays and virtual reality systems X.* Vol. 5006. SPIE, 2003.](https://www.david-romeuf.fr/3D/Anaglyphes/TCAnaglypheLSDubois/ei03.pdf)

```python
from mpl_stereo.example_data import church_left_right
church_left_data, church_right_data = church_left_right

# or
import matplotlib as mpl
church_left_data = mpl.image.imread('church_left.jpg')
church_right_data = mpl.image.imread('church_right.jpg')

axstereo = AxesStereo2D()
axstereo.imshow_stereo([church_left_data, church_right_data])
axstereo.save('church_2d.png', plot_area=True)

axstereo = AxesAnaglyph()
axstereo.imshow_stereo([church_left_data, church_right_data])
axstereo.save('church_anaglyph.png', plot_area=True)
```

<img width="535" height="200" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/church_2d.png">


<img width="265" height="200" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/church_anaglyph.png">

### Wiggle Stereograms
As a final way to show off the stereoscopic effect, we can make a [wiggle stereogram](https://en.wikipedia.org/wiki/Wiggle_stereoscopy) or "wigglegram". This isn't as useful for examining data, but allows seeing the effect without having to train your eyes or using 3D glasses. The sense of depth may be enhanced if you close one eye.

`save()` writes a wiggle animation whenever the filename has an animated extension like `.gif`.

```python
axstereo = AxesStereo2D()  # wiggle also works with AxesStereo3D
axstereo.imshow_stereo([sun_left_data, sun_right_data], cmap='gray')
axstereo.save('sun_wiggle.gif', plot_area=True)  # a .gif animates a wiggle
```

By default the wigglegram rocks between just the two eye views. The `frames` argument samples additional viewpoints across the stereo baseline, producing a smoother rocking motion that oscillates back and forth. For plotted 2D and 3D data, any number of intermediate viewpoints can be synthesized. For images you can supply more than two frames to `imshow_stereo` (ordered left-eye to right-eye) and animate all of them. The first and last are used for the static side-by-side and anaglyph views, while `frames=None` wiggles through every supplied image.

## Advanced Usage

### Working With Plots
The figure and subplot axes can be accessed with the following:
```python
axstereo.fig
axstereo.axs  # (ax_left, ax_right), for AxesStereo2D and AxesStereo3D
axstereo.ax  # for AxesAnaglyph
```

Use `save()` to write the stereogram to file. Whether the result is a static image or animated wigglegrams is chosen from the file extension (e.g. `.png` saves a static image, `.gif` animates a wiggle), or forced with `animate=True`/`animate=False`. Pass `plot_area=True` to save just the plot area without axis decorations. You can also reach the underlying matplotlib figure directly at `axstereo.fig`.

```python
axstereo.save('stereogram.png')              # static image
axstereo.save('stereogram.gif')              # animated wigglegram
axstereo.save('stereogram.png', plot_area=True)  # just the plot area
```

Calling any method on `axstereo` will pass that method call onto all the subplot axes. In the 2D cases, the plotting methods which take in `x` and `y` arguments are intercepted and the additional `z` data is processed to obtain appropriate horizontal offsets.

```python
# For example instead of:
for ax in axstereo.axs:
    ax.set_xlabel('X Label')

# You can do the equivalent:
axstereo.set_xlabel('X Label')

# When applicable, the return values from the two method calls are returned as a tuple
(line_left, line_right) = axstereo.plot(x, y, z)
```

### Parallel vs Cross-Eyed Viewing
By default, the stereograms are set up for the "parallel" / "divergent" / "wall-eyed" viewing method. For "cross-eyed" viewing, initialize with a negative `ipd` parameter. An ipd (Inter-Pupilary Distance) of 65 millimeters is the default, so call `AxesStereo2D(ipd=-65)` for the default cross-eyed viewing.

### Depth and Focal Plane Location
The apparent depth of 2D stereograms can be adjusted with the `zscale` parameter. The default is 1/4th the x-axis range, i.e. `zscale=(max(ax.get_xlim()) - ax.get_xlim()) / 4`. Decrease this number to flatten the stereogram, or increase it to exaggerate the depth.

For 2D plots and anaglyphs, the focal plane can be set with the `zzero` parameter. This is the z value that will be "on the page" when viewing the stereogram.

```python
axstereo = AxesAnaglyph(zzero=min(z)) # data will float above the page
axstereo = AxesAnaglyph(zzero=None) # page will be at the midpoint of the data range (default)
axstereo = AxesAnaglyph(zzero=max(z)) # data will sink below the page
```

<img width="750" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_anaglyph_zzero.png">

### Stereo Squares
The side-by-side, anaglyph, and wiggle plots all have somewhat different viewing methods, and people prefer different methods. For sharing stereograms widely, the `StereoSquare2D` and `StereoSquare3D` classes can be used to generate a single plot that shows all three plots on a 2x2 grid.

```python
from mpl_stereo import StereoSquare2D  # or StereoSquare3D
stereosquare = StereoSquare2D()
stereosquare.imshow_stereo([church_left_data, church_right_data])
# stereosquare.imshow_stereo() is the only special method
# For other plotting call them directly on the object, eg stereosquare.plot(x, y, z)
stereosquare.save('church_2d_square.gif', plot_area=True)
```

```python
stereosquare = StereoSquare3D()
stereosquare.plot(x, y, z, c="k", alpha=0.2)
stereosquare.scatter(x, y, z, c=z, cmap="viridis", s=10)
# We set more frames and a smaller interval for a smoother wiggle
stereosquare.wiggle("trefoil_3d_square.gif", interval=50, frames=5, dpi=100)
```

### Animations
See an example of how to use this with matplotlib animations in [docs/gen_graphics.py](docs/gen_graphics.py).

<img width="500" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_2d_animation.gif">


<img width="500" height="250" src="https://raw.githubusercontent.com/scottshambaugh/mpl_stereo/main/docs/trefoil_3d_animation.gif">

## Viewing Stereograms

These are not [*auto*stereograms](https://en.wikipedia.org/wiki/Autostereogram), like the "Magic Eye" books that were popular in the 1990's. However, they use the same viewing technique. Below is a great video by Vox on how to view stereograms, click on the image and skip to [0:35](https://www.youtube.com/watch?v=v8O8Em_RPNg&t=35s) for the how-to portion and then [4:33](https://www.youtube.com/watch?v=v8O8Em_RPNg&t=4m33s) for more examples.

There is also a great how-to guide at the stereoscopy blog here: [Learning to Free-View: See Stereoscopic Images with the Naked Eye](https://stereoscopy.blog/2022/03/11/learning-to-free-view-see-stereoscopic-images-with-the-naked-eye/)

## Derivation of Geometry
Two eyes with separation `IPD` are looking at a point a distance `z` offset from a focal plane at distance `d`, resulting in view angle `θ`. If this point were projected back to the focal plane, it would be offset by `δ` from where it visually appears on that plane. This offset `δ` is used to displace each point in the 2D stereograms and anaglyphs for each eye based on its `z` value to achieve the stereoscopic effect. The `eye_balance` parameter allocates the total relative displacement of `2δ` between the two eyes.

For 3D stereograms, the azimuth view angles for the two plots are simply shifted by `90 - θ` degrees, with no modification to the data. For a notional `IPD = 65 mm` and distance to screen `d = 350 mm`, `θ ≈ 84.7 deg`, so the plots will be offset by a total of `2 * 6.3 = 12.6 deg`.

```
θ = arctan((d - z) / (IPD / 2))
D / d = (IPD / 2) / (d - z)
δ = D - IPD / 2
  = IPD / 2 * (d / (d - z) - 1)
  = IPD / 2 * z / (d - z)
```

mpl_stereo

mpl_stereo 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

mpl_stereo

mpl_stereo 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README