Upgrade to genjax 0.2.0; incorporate Nishad's fixes for b3d tutorial

bayes3d/__init__.py

@@ -12,7 +12,7 @@
 from .transforms_3d import *
 from .viz import *
 
-RENDERER = None
+RENDERER: "Renderer" = None
 
 __version__ = metadata.version("bayes3d")
 

bayes3d/genjax/model.py

@@ -7,6 +7,7 @@
 from genjax.incremental import Diff, NoChange, UnknownChange
 
 import bayes3d as b
+import bayes3d.scene_graph
 
 from .genjax_distributions import (
     contact_params_uniform,
@@ -127,14 +128,14 @@ def get_far_plane(trace):
 
 
 def add_object(trace, key, obj_id, parent, face_parent, face_child):
-    N = b.get_indices(trace).shape[0] + 1
+    N = get_indices(trace).shape[0] + 1
     choices = trace.get_choices()
     choices[f"parent_{N-1}"] = parent
     choices[f"id_{N-1}"] = obj_id
     choices[f"face_parent_{N-1}"] = face_parent
     choices[f"face_child_{N-1}"] = face_child
     choices[f"contact_params_{N-1}"] = jnp.zeros(3)
-    return model.importance(key, choices, (jnp.arange(N), *trace.get_args()[1:]))[1]
+    return model.importance(key, choices, (jnp.arange(N), *trace.get_args()[1:]))[0]
 
 
 add_object_jit = jax.jit(add_object)
@@ -151,7 +152,7 @@ def print_trace(trace):
 
 
 def viz_trace_meshcat(trace, colors=None):
-    b.clear()
+    b.clear_visualizer()
     b.show_cloud(
         "1", b.apply_transform_jit(trace["image"].reshape(-1, 3), trace["camera_pose"])
     )
@@ -223,14 +224,14 @@ def enumerator(trace, key, *args):
             key,
             chm_builder(addresses, args, chm_args),
             argdiff_f(trace),
-        )[2]
+        )[0]
 
     def enumerator_with_weight(trace, key, *args):
         return trace.update(
             key,
             chm_builder(addresses, args, chm_args),
             argdiff_f(trace),
-        )[1:3]
+        )[0:2]
 
     def enumerator_score(trace, key, *args):
         return enumerator(trace, key, *args).get_score()
@@ -301,4 +302,4 @@ def update_address(trace, key, address, value):
         key,
         genjax.choice_map({address: value}),
         tuple(map(lambda v: Diff(v, UnknownChange), trace.args)),
-    )[2]
+    )[0]

bayes3d/scene_graph.py

@@ -211,16 +211,16 @@ def relative_pose_from_edge(
     face_child,
     dims_child,
 ):
-    x, y, angle = contact_params
-    contact_transform = t3d.transform_from_pos(jnp.array([x, y, 0.0])).dot(
-        t3d.transform_from_axis_angle(jnp.array([1.0, 1.0, 0.0]), jnp.pi).dot(
-            t3d.transform_from_axis_angle(jnp.array([0.0, 0.0, 1.0]), angle)
-        )
-    )
+    contact_transform = contact_params_to_pose(contact_params)
     child_plane = get_contact_planes(dims_child)[face_child]
     return contact_transform.dot(jnp.linalg.inv(child_plane))
 
 
+def relative_pose_from_edge_pose(contact_pose, face_child, dims_child):
+    child_plane = get_contact_planes(dims_child)[face_child]
+    return contact_pose.dot(jnp.linalg.inv(child_plane))
+
+
 relative_pose_from_edge_jit = jax.jit(relative_pose_from_edge)
 relative_pose_from_edge_parallel_jit = jax.jit(
     jax.vmap(

bayes3d/viz/viz.py

@@ -66,11 +66,11 @@ def get_depth_image(image, min_val=None, max_val=None, remove_max=True):
         depth = np.array(image)
 
     if max_val is None:
-        max_val = depth.max()
+        max_val = depth[depth < depth.max()].max(initial=0.0)
     if not remove_max:
         max_val += 1
     if min_val is None:
-        min_val = depth.min()
+        min_val = depth[depth > depth.min()].min(initial=0.0)
 
     mask = (depth < max_val) * (depth > min_val)
     depth[np.logical_not(mask)] = np.nan
@@ -208,7 +208,7 @@ def scale_image(img, factor):
     return img.resize((int(w * factor), int(h * factor)))
 
 
-def vstack_images(images, border=10):
+def vstack_images(images, border=10, bg_color=(255, 255, 255)):
     """Stack images vertically.
 
     Args:
@@ -224,7 +224,7 @@ def vstack_images(images, border=10):
         max_w = max(max_w, w)
         sum_h += h
 
-    full_image = Image.new("RGB", (max_w, sum_h), (255, 255, 255))
+    full_image = Image.new("RGB", (max_w, sum_h), bg_color)
     running_h = 0
     for img in images:
         w, h = img.size
@@ -233,7 +233,7 @@ def vstack_images(images, border=10):
     return full_image
 
 
-def hstack_images(images, border=10):
+def hstack_images(images, border=10, bg_color=(255, 255, 255)):
     """Stack images horizontally.
 
     Args:
@@ -249,7 +249,7 @@ def hstack_images(images, border=10):
         max_h = max(max_h, h)
         sum_w += w
 
-    full_image = Image.new("RGB", (sum_w, max_h), (255, 255, 255))
+    full_image = Image.new("RGB", (sum_w, max_h), bg_color)
     running_w = 0
     for img in images:
         w, h = img.size
@@ -258,7 +258,7 @@ def hstack_images(images, border=10):
     return full_image
 
 
-def hvstack_images(images, h, w, border=10):
+def hvstack_images(images, h, w, border=10, bg_color=(255, 255, 255)):
     """Stack images in a grid.
 
     Args:
@@ -274,10 +274,12 @@ def hvstack_images(images, h, w, border=10):
     images_to_vstack = []
 
     for row_idx in range(h):
-        hstacked_row = hstack_images(images[row_idx * w : (row_idx + 1) * w])
+        hstacked_row = hstack_images(
+            images[row_idx * w : (row_idx + 1) * w], border=border, bg_color=bg_color
+        )
         images_to_vstack.append(hstacked_row)
 
-    return vstack_images(images_to_vstack)
+    return vstack_images(images_to_vstack, border=border, bg_color=bg_color)
 
 
 def multi_panel(

pyproject.toml

@@ -26,7 +26,7 @@ classifiers = [
 ]
 dependencies = [
     "distinctipy",
-    "genjax==0.1.1",
+    "genjax>=0.2.0",
     "graphviz",
     "imageio",
     "matplotlib",