doc: added logo (#125)

New pyproximal logo is added, and some numpy types are fixed based on changes introduced in numpy 1.20.0

Author: mrava87

docs/source/_static/pyproximal.png

@@ -77,8 +77,9 @@
 
 # Sphinx project configuration
 templates_path = ['_templates']
-exclude_patterns = ['_build', '**.ipynb_checkpoints']
+exclude_patterns = ["_build", "**.ipynb_checkpoints", "**.ipynb", "**.md5"]
 source_suffix = '.rst'
+
 # The encoding of source files.
 source_encoding = 'utf-8-sig'
 master_doc = 'index'
@@ -102,7 +103,7 @@
 html_last_updated_fmt = '%b %d, %Y'
 html_title = 'PyProximal'
 html_short_title = 'PyProximal'
-html_logo = 'pyproximal.png'
+html_logo = '_static/pyproximal.png'
 html_favicon = 'favicon.ico'
 html_extra_path = []
 pygments_style = 'default'
@@ -116,6 +117,10 @@
 html_theme_options = {
     'logo_only': True,
     'display_version': True,
+    "logo": {
+            "image_light": "pyproximal_b.png",
+            "image_dark": "pyproximal.png",
+    }
 }
 html_context = {
     'menu_links_name': 'Repository',

docs/source/_static/pyproximal_b.png

@@ -49,7 +49,7 @@ def __call__(self, x, tol=1e-8):
             x = x.reshape(self.dims)
             if self.axis == 0:
                 x = x.T
-            c = np.zeros(self.dims[self.otheraxis], dtype=np.bool)
+            c = np.zeros(self.dims[self.otheraxis], dtype=bool)
             for i in range(self.dims[self.otheraxis]):
                 c[i] = not (np.abs(np.sum(x)) - self.radius < tol or np.any(x[i] < 0))
             c = np.all(c)

docs/source/conf.py

@@ -96,7 +96,7 @@
     sigma = 0.04
     l1 = pyproximal.proximal.L21(ndim=2)
     l2 = pyproximal.proximal.L2(Op=Dop, b=d.ravel(), niter=50, warm=True)
-    Gop = sigma * pylops.Gradient(dims=gt.shape, edge=True, kind='forward', dtype=np.complex)
+    Gop = sigma * pylops.Gradient(dims=gt.shape, edge=True, kind='forward', dtype=np.complex128)
 
     L = sigma ** 2 * 8
     tau = .99 / np.sqrt(L)

pyproximal/proximal/Simplex.py

@@ -46,7 +46,7 @@
 # closest class. This is equivalent to solving our cost function and ignoring
 # the term that minimizes the total interface between the sets. As a result
 # our segmentation boundaries will be very crisp.
-ic = np.floor(ig / dcl).astype(np.int)
+ic = np.floor(ig / dcl).astype(np.int64)
 
 ###############################################################################
 # On the other hand, we can choose to get much smoother boundaries if we use