Tex table

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

@@ -5,7 +5,7 @@
 DETECTION_NAMES = ['car', 'truck', 'bus', 'trailer', 'construction_vehicle', 'pedestrian', 'motorcycle', 'bicycle',
                    'traffic_cone', 'barrier']
 PRETTY_DETECTION_NAMES = {'car': 'Car', 'truck': 'Truck', 'bus': 'Bus', 'trailer': 'Trailer',
-                          'construction_vehicle': 'Construction Vehicle', 'pedestrian': 'Pedestrian',
+                          'construction_vehicle': 'Constr. Veh.', 'pedestrian': 'Pedestrian',
                           'motorcycle': 'Motorcycle', 'bicycle': 'Bicycle', 'traffic_cone': 'Traffic Cone',
                           'barrier': 'Barrier'}
 

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

@@ -1,10 +1,11 @@
 # nuScenes dev-kit.
-# Code written by Holger Caesar, 2019.
+# Code written by Holger Caesar and Alex Lang, 2019.
 # Licensed under the Creative Commons [see licence.txt]
 
 from typing import Dict
 
 import numpy as np
+import json
 from matplotlib import pyplot as plt
 
 from nuscenes.eval.detection.utils import boxes_to_sensor
@@ -250,3 +251,64 @@ def summary_plot(md_list: MetricDataList,
     if savepath is not None:
         plt.savefig(savepath)
         plt.close()
+
+
+def detailed_results_table_tex(metrics_path: str, output_path: str) -> None:
+    """
+    Renders a detailed results table in tex.
+    :param metrics_path: path to a serialized DetectionMetrics file.
+    :param output_path: path to the output file.
+    :return:
+    """
+
+    with open(metrics_path, 'r') as f:
+        metrics = json.load(f)
+
+    tex = ''
+    tex += '\\begin{table}[]\n'
+    tex += '\\small\n'
+    tex += '\\begin{tabular}{| c | c | c | c | c | c | c |} \\hline\n'
+    tex += '\\textbf{Class}    &   \\textbf{AP}  &   \\textbf{ATE} &   \\textbf{ASE} & \\textbf{AOE}   & ' \
+           '\\textbf{AVE}   & ' \
+           '\\textbf{AAE}   \\\\ \\hline ' \
+           '\\hline\n'
+    for name in DETECTION_NAMES:
+        ap = metrics['label_aps'][name]['2.0'] * 100
+        ate = metrics['label_tp_errors'][name]['trans_err']
+        ase = metrics['label_tp_errors'][name]['scale_err']
+        aoe = metrics['label_tp_errors'][name]['orient_err']
+        ave = metrics['label_tp_errors'][name]['vel_err']
+        aae = metrics['label_tp_errors'][name]['attr_err']
+        tex_name = PRETTY_DETECTION_NAMES[name]
+        if name == 'traffic_cone':
+            tex += '{}  &   {:.1f}  &   {:.2f}  &   {:.2f}  &   N/A  &   N/A  &   N/A  \\\\ \\hline\n'.format(
+                tex_name, ap, ate, ase)
+        elif name == 'barrier':
+            tex += '{}  &   {:.1f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  &   N/A  &   N/A  \\\\ \\hline\n'.format(
+                tex_name, ap, ate, ase, aoe)
+        else:
+            tex += '{}  &   {:.1f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  \\\\ ' \
+                   '\\hline\n'.format(tex_name, ap, ate, ase, aoe, ave, aae)
+
+    map_ = metrics['mean_ap'] * 100
+    mate = metrics['tp_errors']['trans_err']
+    mase = metrics['tp_errors']['scale_err']
+    maoe = metrics['tp_errors']['orient_err']
+    mave = metrics['tp_errors']['vel_err']
+    maae = metrics['tp_errors']['attr_err']
+    tex += '\\hline {} &   {:.1f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  &   {:.2f}  \\\\ ' \
+           '\\hline\n'.format('\\textbf{Mean}', map_, mate, mase, maoe, mave, maae)
+
+    tex += '\\end{tabular}\n'
+    tex += '\\caption{Detailed detection performance. '
+    tex += 'AP: average precision (\%), ' \
+           'ATE: average translation error ($m$), ' \
+           'ASE: average scale error ($1-IOU$), ' \
+           'AOE: average orientation error (rad.), ' \
+           'AVE: average velocity error ($m/s$), ' \
+           'AAE: average attribute error ($1-acc$). ' \
+           'nuScenes Detection Score (NDS) = {:.1f} \%{}\n'.format(metrics['weighted_sum'] * 100, '}')
+    tex += '\\end{table}\n'
+
+    with open(output_path, 'w') as f:
+        f.write(tex)