Fix/improve rerun visualization, remove matplotlib as an undocumented dep, bump version (#10)

* remove unnecessary dependency on matplotlib. also prevents a break

* use the rerun blueprint from the visualization we have now

* move rerun config to the package folder, and then always log it on
initialization

so we get a well configured viz by default

* improve docs a bit, and change the name for a workflow build

* bump version because we fixed the visualization in this one

Author: mehermvr

.github/workflows/pypi.yaml

@@ -34,6 +34,7 @@ jobs:
           name: artifact-${{ matrix.os }}
           path: ./wheelhouse/*.whl
   pypi:
+    name: Publish wheels to PyPI
     if: github.event_name == 'release'
     needs: [cibuildwheel]
     runs-on: ubuntu-latest

README.md

@@ -128,9 +128,9 @@ You'll find further usage instructions [here](python#usage).
 For instructions on how to build from source, please check [here](/python/README.md#build-from-source).
 
 <details>
-<summary>Please prefer the ROS version over the python version if you can</summary>
+<summary><b>Please prefer the ROS version over the python version if you can</b></summary>
 
-**Please note:** the ROS version is the intended way to use our odometry system on a robot.
+The ROS version is the intended way to use our odometry system on a robot.
 The python version is slower than the ROS version, not on the odometry itself, but on how we read incoming data, i.e. dataloading.
 Without getting into details, if you can, you should prefer using the ROS version.
 We also provide a way to directly inspect and run our odometry on recorded rosbags (see offline mode in [ROS usage](ros#usage)) which has a performance benefit over the python version.

python/config/rko_lio.rbl

@@ -4,7 +4,7 @@ build-backend = "scikit_build_core.build"
 
 [project]
 name = "rko_lio"
-version = "0.0.2"
+version = "0.0.3"
 description = "A Robust Approach for LiDAR-Inertial Odometry Without Sensor-Specific Modelling"
 dependencies = ["numpy", "typer", "pyyaml", "scipy", "tqdm"]
 authors = [{ name = "Meher Malladi", email = "rm.meher97@gmail.com" }]

python/pyproject.toml

@@ -144,6 +144,7 @@ def cli(
 
             rr.init("rko_lio")
             rr.spawn(memory_limit="2GB")
+            rr.log_file_from_path(Path(__file__).parent / "rko_lio.rbl")
 
         except ImportError:
             raise ImportError(

python/rko_lio/cli.py

@@ -244,25 +244,44 @@ def dump_results_to_disk(self, results_dir: Path, run_name: str):
         self.lio.dump_results_to_disk(results_dir, run_name)
 
 
+viridis_ctrl = np.array(
+    [
+        [68, 1, 84],
+        [71, 44, 122],
+        [59, 81, 139],
+        [44, 113, 142],
+        [33, 144, 140],
+        [39, 173, 129],
+        [92, 200, 99],
+        [170, 220, 50],
+        [253, 231, 37],
+    ],
+    dtype=np.uint8,
+)
+
+
 def height_colors_from_points(points: np.ndarray) -> np.ndarray:
     """
     Given Nx3 array of points, return Nx3 array of RGB colors (dtype uint8)
     mapped from the z values using the viridis colormap.
     Colors are uint8 in range [0, 255].
     """
-    import matplotlib.pyplot as plt
-    from matplotlib import cm
-
     z = points[:, 2]
     z_min, z_max = np.percentile(z, 1), np.percentile(z, 99)  # accounts for any noise
     z_clipped = np.clip(z, z_min, z_max)
+
     if z_max == z_min:
         norm_z = np.zeros_like(z)
     else:
         norm_z = (z_clipped - z_min) / (z_max - z_min)
 
-    viridis = cm.get_cmap("viridis")
-    colors_float = viridis(norm_z)[:, :3]  # Nx3 floats in [0,1]
+    idx = norm_z * (len(viridis_ctrl) - 1)
+    idx_low = np.floor(idx).astype(int)
+    idx_high = np.clip(idx_low + 1, 0, len(viridis_ctrl) - 1)
+    alpha = idx - idx_low
+    colors = (
+        (1 - alpha)[:, None] * viridis_ctrl[idx_low]
+        + alpha[:, None] * viridis_ctrl[idx_high]
+    ).astype(np.uint8)
 
-    colors_uint8 = (colors_float * 255).astype(np.uint8)
-    return colors_uint8
+    return colors

python/rko_lio/lio_pipeline.py

@@ -1,7 +1,7 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <?xml-model href="http://download.ros.org/schema/package_format3.xsd" schematypens="http://www.w3.org/2001/XMLSchema"?><package format="3">
   <name>rko_lio</name>
-  <version>0.0.2</version>
+  <version>0.0.3</version>
   <description>A Robust Approach for LiDAR-Inertial Odometry Without Sensor-Specific Modelling</description>
   <maintainer email="rm.meher97@gmail.com">Meher Malladi</maintainer>
   <license>MIT</license>