changed examples to use pylab.imread in place of scipy.misc.imread

Author: Tristan

README.md

@@ -5,17 +5,17 @@ stl_tools
 Python code to generate STL geometry files from plain text, LaTeX code, and 2D numpy arrays (matrices).
 
 This allows for rapid 3D printing of text, rendered equations, or simple digital images.
-Use them for product prototyping, art, cookie cutters, ice cube trays, chocolate molds, (see [this](http://www.makerbot.com/tutorials/making-chocolate-molds/) 
+Use them for product prototyping, art, cookie cutters, ice cube trays, chocolate molds, (see [this](http://www.makerbot.com/tutorials/making-chocolate-molds/)
 to learn how to make a printed object food-safe) or whatever else you can think of.
 
-Some modification may be needed to the STL or printer settings to get certain shapes to print cleanly (ie. thicker base, 
+Some modification may be needed to the STL or printer settings to get certain shapes to print cleanly (ie. thicker base,
 support structures, etc).
 
-Besides printing, these can also be merged into other 3D meshes for many other 
+Besides printing, these can also be merged into other 3D meshes for many other
 possible uses, using programs such as [Blender](http://www.blender.org/).
 
 Also included is a function that can convert raw LaTeX expressions to high
-quality .png images, which allows for simple inclusion of LaTeX equations into 
+quality .png images, which allows for simple inclusion of LaTeX equations into
 non-LaTeX document editors.
 
  A command-line script for converting images to STL (no text yet) is included in the installation for those who do not
@@ -24,14 +24,14 @@ want to write Python code directly.
 ## Requirements:
 - [Python](http://python.org/) 2.7 or higher (Python 3.x not yet tested, but would probably work)
 - [Numpy](http://www.numpy.org/) 1.7 or higher (for array manipulation)
-- [Scipy](http://www.scipy.org/) 0.12 or higher and [PIL](http://www.pythonware.com/products/pil/)(for image reading and filtering functions)
+- [Scipy](http://www.scipy.org/) 0.12 or higher (for resizing and filtering functions)
 - [Matplotlib](http://matplotlib.org/) 1.2.1 or higher (for rendering text and LaTeX to image data)
 
 ## Installation:
-Run `python setup.py build install` to install. 
+Run `python setup.py build install` to install.
 
 This will check for the 3rd party
-dependencies above and install the library. 
+dependencies above and install the library.
 
 This will also install the
 command line script `image2stl` into the `Python/Scripts` directory.
@@ -68,13 +68,12 @@ The next three examples convert logos to STL, using color information to achieve
 Python code:
 
 ```python
-from scipy.misc import imread
+from pylab import imread
 
-A = imread("examples/example_data/NASA.png")
-A = A[:,:,2] + 1.0*A[:,:,0] # Compose elements from RGBA channels to give depth 
-A = gaussian_filter(A, 1) # smoothing
-
-numpy2stl(A, "examples/NASA.stl", scale=0.05, mask_val = 5.)
+A = 256 * imread("examples/example_data/NASA.png") # 0 - 256 (8 bit) scale
+A = A[:,:, 2] + 1.0*A[:,:, 0] # Compose RGBA channels to give depth
+A = gaussian_filter(A, 1)  # smoothing
+numpy2stl(A, "examples/NASA.stl", scale=0.05, mask_val=5.)
 ```
 Equivalent command-line syntax:
 ```bash
@@ -89,8 +88,8 @@ Equivalent command-line syntax:
 Python code:
 
 ```python
-A = imread("examples/example_data/openmdao.png")
-A =  A[:,:,0] + 1.*A[:,:,3] # Compose elements from RGBA  channels to give depth 
+A = 256.*imread("examples/example_data/openmdao.png") # 0 - 256 (8 bit) scale
+A =  A[:,:,0] + 1.*A[:,:,3] # Compose elements from RGBA  channels to give depth
 A = gaussian_filter(A, 2) # smoothing
 
 numpy2stl(A, "examples/OpenMDAO-logo.stl", scale=0.05, mask_val = 1.)
@@ -110,9 +109,9 @@ Source image vs. output geometry:
 
 Finally, this example renders a LaTeX expression into a png image, then converts this image to an STL.
 
-Note that LaTeX expressions which coincidentally contain special ASCII markers (such as `\n` and `\r`) 
+Note that LaTeX expressions which coincidentally contain special ASCII markers (such as `\n` and `\r`)
 have to be escaped with an additional slash in those positions in order to be properly rendered, unless these
-markers are intended. 
+markers are intended.
 
 Python code:
 
@@ -123,13 +122,13 @@ text = ("$\oint_{\Gamma} (A\, dx + B\, dy) = \iint_{U} \left(\\frac{\partial "
         "B}{\partial x} - \\frac{\partial A}{\partial y}\\right)\ dxdy$ \n\n "
         "$\\frac{\partial \\rho}{\partial t} + \\frac{\partial}{\partial x_j}"
         "\left[ \\rho u_j \\right] = 0$")
-text2png(text, "examples/Greens-Theorem_Navier-Stokes", fontsize=50) #save png 
+text2png(text, "examples/Greens-Theorem_Navier-Stokes", fontsize=50) #save png
 
 A = imread("examples/Greens-Theorem_Navier-Stokes.png") # read from rendered png
 A = A.mean(axis=2) #grayscale projection
-A = gaussian_filter(A.max() - A, 1.) 
+A = gaussian_filter(A.max() - A, 1.)
 
-numpy2stl(A, "examples/Greens-Theorem_Navier-Stokes.stl", scale=0.2, 
+numpy2stl(A, "examples/Greens-Theorem_Navier-Stokes.stl", scale=0.2,
                                                          mask_val = 5.)
 ```
 Source image vs. output geometry:
@@ -151,7 +150,7 @@ There are 3 principal functions (no classes) to import and use from stl_tools:
      fn (string) -  filename to use for STL file
 
     Optional input:
-     scale (float)  -  scales the height (surface) of the 
+     scale (float)  -  scales the height (surface) of the
                        resulting STL mesh. Tune to match needs
 
      mask_val (float) - any element of the inputted array that is less
@@ -164,14 +163,14 @@ There are 3 principal functions (no classes) to import and use from stl_tools:
 
      max_width, max_depth, max_height (floats) - maximum size of the stl
                                                 object (in mm). Match this to
-                                                the dimensions of a 3D printer platform 
+                                                the dimensions of a 3D printer platform
 
     Returns: (None)
 
-`numpy2stl()` is the main function of this repository. 
+`numpy2stl()` is the main function of this repository.
 
 It takes a 2D numpy array and output filename
-as input, and writes an STL file. 
+as input, and writes an STL file.
 
 Each element of the array is tesselated to its neighbors to produce 2 triangular faces for
 every 4 contiguous elements. The depth axis of any vertex is taken to be the value of the array corresponding to that point.
@@ -183,7 +182,7 @@ Array elements which are less than this value will not be included as vertices.
 It takes a bit of trial-and-error to get these just right. Plotting the intermediate arrays
 with a colorbar (to show scaling) helps in finding decent values.
 
-The `max_width`, `max_height`, and `max_depth` values are measurements (in mm) used to scale the final output to 
+The `max_width`, `max_height`, and `max_depth` values are measurements (in mm) used to scale the final output to
 the largest size that can fit onto your 3D printer platform. Default values are for the MakerBot Replicator.
 
 
@@ -201,7 +200,7 @@ the largest size that can fit onto your 3D printer platform. Default values are
     Returns: (None)
 
 `text2png()` was written as an intermediate helper function to render text to pngs, to then be imported, filtered, and meshed.
-However, it may be useful in it's own right. For example, it can be used alone to render LaTeX expressions into images, to be imported into WYSIWYG document editors like MS Word or LibreOffice Writer. 
+However, it may be useful in it's own right. For example, it can be used alone to render LaTeX expressions into images, to be imported into WYSIWYG document editors like MS Word or LibreOffice Writer.
 
 ### `stl_tools.text2array`
     text2array(text)
@@ -211,7 +210,7 @@ However, it may be useful in it's own right. For example, it can be used alone t
      text (string) -  text to render
 
     Returns: A (ndarray) - 2D numpy array of rendered text
-    
+
 
 `text2array()` renders inputted text using `text2png()`, but imports the resulting png as an ndarray and deletes the intermediate file.
 There may be a direct way to render the matplotlib figure as an array without using an intermediate file, but I could not seem to find a simple
@@ -231,10 +230,10 @@ arguments. See the examples above for usage.
 For images of standard sizes for modern cameras, the resulting STL filesize can be pretty large.
 
 - Just like was shown in the examples, applying a simple filtering function to smooth
-sharp edges results in an STL geometry that is likely to be more easily printable. Fine tuning in a 
+sharp edges results in an STL geometry that is likely to be more easily printable. Fine tuning in a
 program like Photoshop or Gimp can also help prevent spikes/jagged edges in the geometry.
 
-- To make a proper mold, scale up the edges of the source image to match the maximum pixel value of the image (or higher), to form a lip. 
+- To make a proper mold, scale up the edges of the source image to match the maximum pixel value of the image (or higher), to form a lip.
 For example:
 ```python
 m, n = A.shape
@@ -244,13 +243,22 @@ A[:, 0::n-1] = border_val # make left and right lip
 ```
 In practice, the border may need to be thicker than 1 pixel.
 
+## Known bugs:
+
+- `scipy.misc.imread` has some odd behavior for me on my Linux box (64-bit
+Mint 15). When an image is read, an empty 0-dim array is returned. This is
+supposedly due to a versioning issue with PIL, which I can't seem to work
+around successfully. `pylab.imread` works fine, and seems consistent
+across platforms.
+
+
 ## Todo/future features:
 
-- Photos of actual printed models. 
+- Photos of actual printed models.
 
 - I'm looking into writing a utility function to refine STL meshes by removing redundant vertices (so that wide flat spaces aren't packed with dense tessellations)
 
-- It's possible to apply various warping functions to the resulting 
-meshes. So you could load an image, warp the result into a cylinder, and have a 
+- It's possible to apply various warping functions to the resulting
+meshes. So you could load an image, warp the result into a cylinder, and have a
 textured column (or something like that).
 

examples.py

@@ -1,6 +1,7 @@
-from scipy.misc import lena, imresize, imread
+from scipy.misc import lena, imresize
+from pylab import imread
 from scipy.ndimage import gaussian_filter
-from stl_tools import numpy2stl, text2png, text2array
+from stl_tools import numpy2stl, text2png
 
 
 """
@@ -16,13 +17,13 @@
 A = gaussian_filter(A, 1)  # smoothing
 numpy2stl(A, "examples/Lena.stl", scale=0.1)
 
-A = imread("examples/example_data/NASA.png")
-A = A[:, :, 2] + 1.0*A[:,:, 0] # Compose RGBA channels to give depth 
+A = 256 * imread("examples/example_data/NASA.png")
+A = A[:,:, 2] + 1.0*A[:,:, 0] # Compose RGBA channels to give depth
 A = gaussian_filter(A, 1)  # smoothing
 numpy2stl(A, "examples/NASA.stl", scale=0.05, mask_val=5.)
 
-A = imread("examples/example_data/openmdao.png")
-A =  A[:, :, 0] + 1.*A[:,:, 3] # Compose some elements from RGBA to give depth 
+A = 256 * imread("examples/example_data/openmdao.png")
+A =  A[:,:, 0] + 1.*A[:,:, 3] # Compose some elements from RGBA to give depth
 A = gaussian_filter(A, 2)  # smoothing
 numpy2stl(A, "examples/OpenMDAO-logo.stl", scale=0.05, mask_val=1.)
 
@@ -33,7 +34,7 @@
 # save png
 text2png(text, "examples/Greens-Theorem_Navier-Stokes", fontsize=50)
 # read from rendered png
-A = imread("examples/Greens-Theorem_Navier-Stokes.png")
+A = 256 * imread("examples/Greens-Theorem_Navier-Stokes.png")
 A = A.mean(axis=2)  # grayscale projection
 A = gaussian_filter(A.max() - A, 1.0)
 numpy2stl(A, "examples/Greens-Theorem_Navier-Stokes.stl", scale=0.2,

setup.py

@@ -2,7 +2,7 @@
 
 setup(name='stl_tools',
       version='0.1',
-      install_requires=['numpy', 'pil', 'scipy', 'matplotlib'],
+      install_requires=['numpy', 'scipy', 'matplotlib'],
       description="Generate STL files from numpy arrays and text",
       author='Tristan Hearn',
       author_email='[email protected]',

stl_tools/image2stl.py

@@ -2,7 +2,7 @@
 
 import numpy as np
 from numpy2stl import numpy2stl
-from scipy.misc import imread
+from pylab import imread
 from scipy.ndimage import gaussian_filter
 
 
@@ -32,8 +32,10 @@ def image2stl():
 
     args = vars(parser.parse_args())
 
-    f_args = {f_arg: float(args[f_arg]) for f_arg in _float_args if args[f_arg]}
-    b_args = {b_arg: bool(int(args[b_arg])) for b_arg in _bool_args if args[b_arg]}
+    f_args = {f_arg: float(args[f_arg])
+              for f_arg in _float_args if args[f_arg]}
+    b_args = {b_arg: bool(int(args[b_arg]))
+              for b_arg in _bool_args if args[b_arg]}
 
     kwargs = dict(f_args, **b_args)
 
@@ -42,7 +44,7 @@ def image2stl():
     if not fn:
         fn = '.'.join([src.split('.')[0], "stl"])
 
-    A = imread(src)
+    A = 256. * imread(src)
     L = len(A.shape)
     w = args['RGBA_weights']
     if L > 2: