add hota iou calculation along with detection/association accuracy

motmetrics/metrics.py

@@ -751,6 +751,36 @@ def idf1_m(partials, idtp, num_objects, num_predictions):
     return math_util.quiet_divide(2 * idtp, num_objects + num_predictions)
 
 
+def det_acc(df, num_detections, num_objects, num_false_positives):
+    """Detection accuracy.
+
+    Source: https://jonathonluiten.medium.com/how-to-evaluate-tracking-with-the-hota-metrics-754036d183e1
+    """
+
+    del df  # unused
+    return math_util.quiet_divide(num_detections, num_objects + num_false_positives)
+
+
+def ass_acc(df, idtp, idfn, idfp):
+    """Association accuracy.
+
+    Source: https://jonathonluiten.medium.com/how-to-evaluate-tracking-with-the-hota-metrics-754036d183e1
+    """
+
+    del df  # unused
+    return math_util.quiet_divide(idtp, idtp + idfn + idfp)
+
+
+def hota_iou(df, det_acc, ass_acc):
+    """HOTA metric at a specific iou.
+
+    Source: https://jonathonluiten.medium.com/how-to-evaluate-tracking-with-the-hota-metrics-754036d183e1
+    """
+
+    del df
+    return (det_acc * ass_acc)**0.5
+
+
 for one in simple_add_func:
     name = one.__name__
 
@@ -802,6 +832,10 @@ def create():
     m.register(idr, formatter="{:.1%}".format)
     m.register(idf1, formatter="{:.1%}".format)
 
+    m.register(det_acc, formatter="{:.1%}".format)
+    m.register(ass_acc, formatter="{:.1%}".format)
+    m.register(hota_iou, formatter="{:.1%}".format)
+
     return m
 
 

motmetrics/tests/test_metrics.py

@@ -15,6 +15,7 @@
 
 import numpy as np
 import pandas as pd
+import pytest
 from pytest import approx
 
 import motmetrics as mm
@@ -343,6 +344,42 @@ def test_mota_motp():
     assert metr["num_frames"] == 6
 
 
+def test_hota():
+    """Tests values of HOTA and its dependents."""
+    acc = mm.MOTAccumulator()
+
+    # All FP
+    acc.update([], [1, 2], [], frameid=0)
+    # All miss
+    acc.update([1, 2], [], [], frameid=1)
+    # Match
+    acc.update([1, 2], [1, 2], [[1, 0.5], [0.3, 1]], frameid=2)
+    # Switch
+    acc.update([1, 2], [1, 2], [[0.2, np.nan], [np.nan, 0.1]], frameid=3)
+    # Match. Better new match is available but should prefer history
+    acc.update([1, 2], [1, 2], [[5, 1], [1, 5]], frameid=4)
+    # No data
+    acc.update([], [], [], frameid=5)
+
+    mh = mm.metrics.create()
+    metr = mh.compute(
+        acc,
+        return_dataframe=False,
+        return_cached=True,
+        metrics=[
+            "hota_iou",
+            "det_acc",
+            "ass_acc",
+            "num_frames",
+        ],
+    )
+
+    assert metr["det_acc"] == approx(6 / (8 + 2))
+    assert metr["ass_acc"] == approx(6 / (8 + 2))
+    assert metr["hota_iou"] == approx(0.6)
+    assert metr["num_frames"] == 6
+
+
 def test_ids():
     """Test metrics with frame IDs specified manually."""
     acc = mm.MOTAccumulator()