Miscellaneous improvements: Plot settings, default number of sweeps (n…

…utonomy#215)

* Don't include pycache folders in pip package, added new author

* Include license in pip package

* Improve evaluation plot limits

* Use known stable Docker build

* Fixed deprecated link

* Changed default number of sweeps to 5, cleanup

* added typing

README.md

@@ -45,7 +45,7 @@ To install this expansion, please follow these steps:
 
 ## Devkit setup
 The devkit is tested for Python 3.6 and Python 3.7.
-To install Python, please check [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/installation.md#install-python).
+To install Python, please check [here](https://github.com/nutonomy/nuscenes-devkit/blob/master/setup/installation.md#install-python).
 
 Our devkit is available and can be installed via [pip](https://pip.pypa.io/en/stable/installing/) :
 ```

python-sdk/nuscenes/eval/detection/render.py

@@ -186,7 +186,9 @@ def class_tp_curve(md_list: MetricDataList,
     md = md_list[(detection_name, dist_th_tp)]
     min_recall_ind = round(100 * min_recall)
     if min_recall_ind <= md.max_recall_ind:
-        ylimit = max([max(getattr(md, metric)[min_recall_ind:md.max_recall_ind + 1]) for metric in TP_METRICS]) * 1.1
+        # For traffic_cone and barrier only a subset of the metrics are plotted.
+        rel_metrics = [m for m in TP_METRICS if not np.isnan(metrics.get_label_tp(detection_name, m))]
+        ylimit = max([max(getattr(md, metric)[min_recall_ind:md.max_recall_ind + 1]) for metric in rel_metrics]) * 1.1
     else:
         ylimit = 1.0
 

python-sdk/nuscenes/utils/data_classes.py

@@ -59,37 +59,36 @@ def from_file_multisweep(cls,
                              sample_rec: Dict,
                              chan: str,
                              ref_chan: str,
-                             nsweeps: int = 26,
+                             nsweeps: int = 5,
                              min_distance: float = 1.0) -> Tuple['PointCloud', np.ndarray]:
         """
         Return a point cloud that aggregates multiple sweeps.
         As every sweep is in a different coordinate frame, we need to map the coordinates to a single reference frame.
         As every sweep has a different timestamp, we need to account for that in the transformations and timestamps.
         :param nusc: A NuScenes instance.
         :param sample_rec: The current sample.
-        :param chan: The radar channel from which we track back n sweeps to aggregate the point cloud.
+        :param chan: The lidar/radar channel from which we track back n sweeps to aggregate the point cloud.
         :param ref_chan: The reference channel of the current sample_rec that the point clouds are mapped to.
         :param nsweeps: Number of sweeps to aggregated.
         :param min_distance: Distance below which points are discarded.
         :return: (all_pc, all_times). The aggregated point cloud and timestamps.
         """
-
-        # Init
+        # Init.
         points = np.zeros((cls.nbr_dims(), 0))
         all_pc = cls(points)
         all_times = np.zeros((1, 0))
 
-        # Get reference pose and timestamp
+        # Get reference pose and timestamp.
         ref_sd_token = sample_rec['data'][ref_chan]
         ref_sd_rec = nusc.get('sample_data', ref_sd_token)
         ref_pose_rec = nusc.get('ego_pose', ref_sd_rec['ego_pose_token'])
         ref_cs_rec = nusc.get('calibrated_sensor', ref_sd_rec['calibrated_sensor_token'])
         ref_time = 1e-6 * ref_sd_rec['timestamp']
 
-        # Homogeneous transform from ego car frame to reference frame
+        # Homogeneous transform from ego car frame to reference frame.
         ref_from_car = transform_matrix(ref_cs_rec['translation'], Quaternion(ref_cs_rec['rotation']), inverse=True)
 
-        # Homogeneous transformation matrix from global to _current_ ego car frame
+        # Homogeneous transformation matrix from global to _current_ ego car frame.
         car_from_global = transform_matrix(ref_pose_rec['translation'], Quaternion(ref_pose_rec['rotation']),
                                            inverse=True)
 
@@ -116,7 +115,7 @@ def from_file_multisweep(cls,
 
             # Remove close points and add timevector.
             current_pc.remove_close(min_distance)
-            time_lag = ref_time - 1e-6 * current_sd_rec['timestamp']  # positive difference
+            time_lag = ref_time - 1e-6 * current_sd_rec['timestamp']  # Positive difference.
             times = time_lag * np.ones((1, current_pc.nbr_points()))
             all_times = np.hstack((all_times, times))
 
@@ -182,31 +181,31 @@ def transform(self, transf_matrix: np.ndarray) -> None:
     def render_height(self,
                       ax: Axes,
                       view: np.ndarray = np.eye(4),
-                      x_lim: Tuple = (-20, 20),
-                      y_lim: Tuple = (-20, 20),
+                      x_lim: Tuple[float, float] = (-20, 20),
+                      y_lim: Tuple[float, float] = (-20, 20),
                       marker_size: float = 1) -> None:
         """
         Very simple method that applies a transformation and then scatter plots the points colored by height (z-value).
         :param ax: Axes on which to render the points.
         :param view: <np.float: n, n>. Defines an arbitrary projection (n <= 4).
-        :param x_lim: (min <float>, max <float>). x range for plotting.
-        :param y_lim: (min <float>, max <float>). y range for plotting.
+        :param x_lim: (min, max). x range for plotting.
+        :param y_lim: (min, max). y range for plotting.
         :param marker_size: Marker size.
         """
         self._render_helper(2, ax, view, x_lim, y_lim, marker_size)
 
     def render_intensity(self,
                          ax: Axes,
                          view: np.ndarray = np.eye(4),
-                         x_lim: Tuple = (-20, 20),
-                         y_lim: Tuple = (-20, 20),
+                         x_lim: Tuple[float, float] = (-20, 20),
+                         y_lim: Tuple[float, float] = (-20, 20),
                          marker_size: float = 1) -> None:
         """
         Very simple method that applies a transformation and then scatter plots the points colored by intensity.
         :param ax: Axes on which to render the points.
         :param view: <np.float: n, n>. Defines an arbitrary projection (n <= 4).
-        :param x_lim: (min <float>, max <float>).
-        :param y_lim: (min <float>, max <float>).
+        :param x_lim: (min, max).
+        :param y_lim: (min, max).
         :param marker_size: Marker size.
         """
         self._render_helper(3, ax, view, x_lim, y_lim, marker_size)
@@ -215,16 +214,16 @@ def _render_helper(self,
                        color_channel: int,
                        ax: Axes,
                        view: np.ndarray,
-                       x_lim: Tuple,
-                       y_lim: Tuple,
+                       x_lim: Tuple[float, float],
+                       y_lim: Tuple[float, float],
                        marker_size: float) -> None:
         """
         Helper function for rendering.
         :param color_channel: Point channel to use as color.
         :param ax: Axes on which to render the points.
         :param view: <np.float: n, n>. Defines an arbitrary projection (n <= 4).
-        :param x_lim: (min <float>, max <float>).
-        :param y_lim: (min <float>, max <float>).
+        :param x_lim: (min, max).
+        :param y_lim: (min, max).
         :param marker_size: Marker size.
         """
         points = view_points(self.points[:3, :], view, normalize=False)
@@ -555,7 +554,7 @@ def render(self,
                view: np.ndarray = np.eye(3),
                normalize: bool = False,
                colors: Tuple = ('b', 'r', 'k'),
-               linewidth: float = 2):
+               linewidth: float = 2) -> None:
         """
         Renders the box in the provided Matplotlib axis.
         :param axis: Axis onto which the box should be drawn.

setup/setup.py

@@ -26,11 +26,14 @@ def get_dirlist(_rootdir):
 rootdir = 'python-sdk'
 packages = [d.replace('/', '.').replace('{}.'.format(rootdir), '') for d in get_dirlist(rootdir)]
 
+# Filter out Python cache folders
+packages = [p for p in packages if not p.endswith('__pycache__')]
+
 setuptools.setup(
     name='nuscenes-devkit',
     version='1.0.2',
     author='Holger Caesar, Oscar Beijbom, Qiang Xu, Varun Bankiti, Alex H. Lang, Sourabh Vora, Venice Erin Liong, '
-           'Chris Li, Sergi Widjaja et al.',
+           'Chris Li, Sergi Widjaja, Kiwoo Shin et al.',
     author_email='nuscenes@nutonomy.com',
     description='The official devkit of the nuScenes dataset (www.nuscenes.org).',
     long_description=long_description,
@@ -45,5 +48,7 @@ def get_dirlist(_rootdir):
     classifiers=[
         'Programming Language :: Python :: 3.6',
         'Operating System :: OS Independent',
+        'License :: Free for non-commercial use'
     ],
+    license='cc-by-nc-sa-4.0'
 )