diff --git a/media/RESULTS.md b/media/RESULTS.md
index 823c8c3..632ed24 100644
--- a/media/RESULTS.md
+++ b/media/RESULTS.md
@@ -286,16 +286,16 @@ Results of the model in various experiments on different datasets.
   "person_nums": {
     "total_frames": 301,
     "total_labels": 477,
-    "total_preds": 888,
+    "total_preds": 814,
     "considered_empty": 0,
     "valid_preds": 477,
-    "invalid_preds": 411,
+    "invalid_preds": 337,
     "missing": 0,
-    "invalid_fraction": 0.46284,
-    "precision": 0.53716,
+    "invalid_fraction": 0.414,
+    "precision": 0.586,
     "recall": 1.0,
-    "f1": 0.6989,
-    "non_empty": 888
+    "f1": 0.73896,
+    "non_empty": 814
   },
   "mpjpe": {
     "count": 477,
@@ -313,11 +313,11 @@ Results of the model in various experiments on different datasets.
     "recall-0.5": 1.0,
     "num_labels": 477,
     "ap-0.025": 0.0,
-    "ap-0.05": 0.340331,
-    "ap-0.1": 0.693196,
-    "ap-0.15": 0.704408,
-    "ap-0.25": 0.704408,
-    "ap-0.5": 0.704408
+    "ap-0.05": 0.345013,
+    "ap-0.1": 0.702867,
+    "ap-0.15": 0.714372,
+    "ap-0.25": 0.714372,
+    "ap-0.5": 0.714372
   },
   "head": {
     "count": 477,
@@ -1569,56 +1569,56 @@ Results of the model in various experiments on different datasets.
 (10 cameras)
 ```json
 {
-  "avg_time_2d": 0.20488240776992425,
-  "avg_time_3d": 0.0016675780459148128,
-  "avg_fps": 4.841443082410108
+  "avg_time_2d": 0.20779247690991656,
+  "avg_time_3d": 0.0016487220438515268,
+  "avg_fps": 4.774609794994247
 }
 {
   "person_nums": {
     "total_frames": 420,
     "total_labels": 1466,
-    "total_preds": 1735,
+    "total_preds": 1527,
     "considered_empty": 0,
     "valid_preds": 1465,
-    "invalid_preds": 270,
+    "invalid_preds": 62,
     "missing": 1,
-    "invalid_fraction": 0.15562,
-    "precision": 0.84438,
+    "invalid_fraction": 0.0406,
+    "precision": 0.9594,
     "recall": 0.99932,
-    "f1": 0.91534,
-    "non_empty": 1735
+    "f1": 0.97895,
+    "non_empty": 1527
   },
   "mpjpe": {
     "count": 1465,
-    "mean": 0.037085,
-    "median": 0.032344,
-    "std": 0.017223,
-    "sem": 0.00045,
+    "mean": 0.037082,
+    "median": 0.032321,
+    "std": 0.017242,
+    "sem": 0.000451,
     "min": 0.013848,
     "max": 0.136363,
-    "recall-0.025": 0.186903,
-    "recall-0.05": 0.85266,
+    "recall-0.025": 0.188267,
+    "recall-0.05": 0.851978,
     "recall-0.1": 0.989086,
     "recall-0.15": 0.999318,
     "recall-0.25": 0.999318,
     "recall-0.5": 0.999318,
     "num_labels": 1466,
-    "ap-0.025": 0.085997,
-    "ap-0.05": 0.802543,
-    "ap-0.1": 0.967268,
-    "ap-0.15": 0.978621,
-    "ap-0.25": 0.978621,
-    "ap-0.5": 0.978621
+    "ap-0.025": 0.087846,
+    "ap-0.05": 0.807698,
+    "ap-0.1": 0.975275,
+    "ap-0.15": 0.986743,
+    "ap-0.25": 0.986743,
+    "ap-0.5": 0.986743
   },
   "nose": {
     "count": 1462,
-    "mean": 0.012376,
-    "median": 0.010938,
-    "std": 0.007779,
+    "mean": 0.012384,
+    "median": 0.010912,
+    "std": 0.007797,
     "sem": 0.000204,
     "min": 0.001275,
     "max": 0.124831,
-    "recall-0.025": 0.957621,
+    "recall-0.025": 0.956254,
     "recall-0.05": 0.993165,
     "recall-0.1": 0.998633,
     "recall-0.15": 0.999316,
@@ -1628,13 +1628,13 @@ Results of the model in various experiments on different datasets.
   },
   "shoulder_left": {
     "count": 1465,
-    "mean": 0.018742,
-    "median": 0.016661,
-    "std": 0.00976,
-    "sem": 0.000255,
+    "mean": 0.018746,
+    "median": 0.016641,
+    "std": 0.009717,
+    "sem": 0.000254,
     "min": 0.001103,
     "max": 0.11138,
-    "recall-0.025": 0.800136,
+    "recall-0.025": 0.798772,
     "recall-0.05": 0.991132,
     "recall-0.1": 0.998636,
     "recall-0.15": 0.999318,
@@ -1644,13 +1644,13 @@ Results of the model in various experiments on different datasets.
   },
   "shoulder_right": {
     "count": 1464,
-    "mean": 0.02051,
+    "mean": 0.020464,
     "median": 0.019186,
-    "std": 0.008943,
-    "sem": 0.000234,
+    "std": 0.008812,
+    "sem": 0.00023,
     "min": 0.002013,
-    "max": 0.093441,
-    "recall-0.025": 0.768601,
+    "max": 0.087963,
+    "recall-0.025": 0.769966,
     "recall-0.05": 0.990444,
     "recall-0.1": 0.999317,
     "recall-0.15": 0.999317,
@@ -1660,15 +1660,15 @@ Results of the model in various experiments on different datasets.
   },
   "elbow_left": {
     "count": 1464,
-    "mean": 0.020864,
-    "median": 0.018441,
-    "std": 0.011938,
-    "sem": 0.000312,
+    "mean": 0.021041,
+    "median": 0.018392,
+    "std": 0.012619,
+    "sem": 0.00033,
     "min": 0.002104,
-    "max": 0.095363,
-    "recall-0.025": 0.711945,
-    "recall-0.05": 0.974744,
-    "recall-0.1": 0.999317,
+    "max": 0.147682,
+    "recall-0.025": 0.709215,
+    "recall-0.05": 0.972696,
+    "recall-0.1": 0.998635,
     "recall-0.15": 0.999317,
     "recall-0.25": 0.999317,
     "recall-0.5": 0.999317,
@@ -1676,14 +1676,14 @@ Results of the model in various experiments on different datasets.
   },
   "elbow_right": {
     "count": 1462,
-    "mean": 0.019564,
-    "median": 0.017384,
-    "std": 0.011529,
-    "sem": 0.000302,
+    "mean": 0.019468,
+    "median": 0.017347,
+    "std": 0.011368,
+    "sem": 0.000297,
     "min": 0.001927,
     "max": 0.132247,
-    "recall-0.025": 0.802461,
-    "recall-0.05": 0.970608,
+    "recall-0.025": 0.805878,
+    "recall-0.05": 0.971975,
     "recall-0.1": 0.997949,
     "recall-0.15": 0.999316,
     "recall-0.25": 0.999316,
@@ -1692,13 +1692,13 @@ Results of the model in various experiments on different datasets.
   },
   "wrist_left": {
     "count": 1433,
-    "mean": 0.020859,
+    "mean": 0.020901,
     "median": 0.015088,
-    "std": 0.021069,
-    "sem": 0.000557,
+    "std": 0.021121,
+    "sem": 0.000558,
     "min": 0.001506,
     "max": 0.194344,
-    "recall-0.025": 0.767085,
+    "recall-0.025": 0.763598,
     "recall-0.05": 0.937936,
     "recall-0.1": 0.982566,
     "recall-0.15": 0.994421,
@@ -1708,14 +1708,14 @@ Results of the model in various experiments on different datasets.
   },
   "wrist_right": {
     "count": 1456,
-    "mean": 0.020392,
-    "median": 0.0137,
-    "std": 0.020859,
-    "sem": 0.000547,
+    "mean": 0.020303,
+    "median": 0.013717,
+    "std": 0.020686,
+    "sem": 0.000542,
     "min": 0.000284,
     "max": 0.212342,
-    "recall-0.025": 0.770604,
-    "recall-0.05": 0.933379,
+    "recall-0.025": 0.773352,
+    "recall-0.05": 0.934753,
     "recall-0.1": 0.984203,
     "recall-0.15": 0.997253,
     "recall-0.25": 1.0,
@@ -1724,15 +1724,15 @@ Results of the model in various experiments on different datasets.
   },
   "hip_left": {
     "count": 1464,
-    "mean": 0.050256,
-    "median": 0.048397,
-    "std": 0.01907,
-    "sem": 0.000499,
+    "mean": 0.050265,
+    "median": 0.048338,
+    "std": 0.019008,
+    "sem": 0.000497,
     "min": 0.008094,
     "max": 0.148516,
-    "recall-0.025": 0.049829,
-    "recall-0.05": 0.546075,
-    "recall-0.1": 0.975427,
+    "recall-0.025": 0.049147,
+    "recall-0.05": 0.54744,
+    "recall-0.1": 0.976109,
     "recall-0.15": 0.999317,
     "recall-0.25": 0.999317,
     "recall-0.5": 0.999317,
@@ -1740,14 +1740,14 @@ Results of the model in various experiments on different datasets.
   },
   "hip_right": {
     "count": 1465,
-    "mean": 0.050045,
-    "median": 0.048856,
-    "std": 0.016859,
-    "sem": 0.000441,
+    "mean": 0.050041,
+    "median": 0.048863,
+    "std": 0.016842,
+    "sem": 0.00044,
     "min": 0.007258,
     "max": 0.138747,
-    "recall-0.025": 0.051842,
-    "recall-0.05": 0.521146,
+    "recall-0.025": 0.05116,
+    "recall-0.05": 0.520464,
     "recall-0.1": 0.988404,
     "recall-0.15": 0.999318,
     "recall-0.25": 0.999318,
@@ -1756,13 +1756,13 @@ Results of the model in various experiments on different datasets.
   },
   "knee_left": {
     "count": 1464,
-    "mean": 0.038364,
-    "median": 0.032286,
-    "std": 0.027217,
-    "sem": 0.000712,
+    "mean": 0.038353,
+    "median": 0.032352,
+    "std": 0.027187,
+    "sem": 0.000711,
     "min": 0.002051,
     "max": 0.275419,
-    "recall-0.025": 0.333106,
+    "recall-0.025": 0.331058,
     "recall-0.05": 0.759727,
     "recall-0.1": 0.970648,
     "recall-0.15": 0.990444,
@@ -1772,59 +1772,59 @@ Results of the model in various experiments on different datasets.
   },
   "knee_right": {
     "count": 1458,
-    "mean": 0.041819,
-    "median": 0.035897,
-    "std": 0.026627,
-    "sem": 0.000698,
+    "mean": 0.041659,
+    "median": 0.035863,
+    "std": 0.026298,
+    "sem": 0.000689,
     "min": 0.004598,
     "max": 0.242773,
-    "recall-0.025": 0.242632,
-    "recall-0.05": 0.731323,
-    "recall-0.1": 0.963674,
-    "recall-0.15": 0.989034,
+    "recall-0.025": 0.243317,
+    "recall-0.05": 0.732008,
+    "recall-0.1": 0.964359,
+    "recall-0.15": 0.990404,
     "recall-0.25": 0.999315,
     "recall-0.5": 0.999315,
     "num_labels": 1459
   },
   "ankle_left": {
     "count": 1457,
-    "mean": 0.085272,
-    "median": 0.043075,
-    "std": 0.101395,
+    "mean": 0.085318,
+    "median": 0.04279,
+    "std": 0.101384,
     "sem": 0.002657,
     "min": 0.000814,
     "max": 0.494931,
-    "recall-0.025": 0.346548,
-    "recall-0.05": 0.545455,
-    "recall-0.1": 0.712919,
-    "recall-0.15": 0.828435,
+    "recall-0.025": 0.347915,
+    "recall-0.05": 0.546138,
+    "recall-0.1": 0.712235,
+    "recall-0.15": 0.827751,
     "recall-0.25": 0.917293,
     "recall-0.5": 0.995899,
     "num_labels": 1463
   },
   "ankle_right": {
     "count": 1447,
-    "mean": 0.077062,
-    "median": 0.042773,
-    "std": 0.096671,
-    "sem": 0.002542,
+    "mean": 0.077143,
+    "median": 0.042829,
+    "std": 0.096759,
+    "sem": 0.002545,
     "min": 0.001957,
     "max": 0.49866,
-    "recall-0.025": 0.315753,
+    "recall-0.025": 0.315068,
     "recall-0.05": 0.560274,
-    "recall-0.1": 0.793151,
-    "recall-0.15": 0.850685,
-    "recall-0.25": 0.908219,
+    "recall-0.1": 0.791781,
+    "recall-0.15": 0.85137,
+    "recall-0.25": 0.908904,
     "recall-0.5": 0.991096,
     "num_labels": 1460
   },
   "joint_recalls": {
     "num_labels": 18990,
-    "recall-0.025": 0.53133,
-    "recall-0.05": 0.80379,
-    "recall-0.1": 0.95082,
+    "recall-0.025": 0.53102,
+    "recall-0.05": 0.80395,
+    "recall-0.1": 0.95061,
     "recall-0.15": 0.97241,
-    "recall-0.25": 0.98562,
+    "recall-0.25": 0.98568,
     "recall-0.5": 0.99816
   }
 }
@@ -2119,23 +2119,23 @@ Results of the model in various experiments on different datasets.
   "person_nums": {
     "total_frames": 210,
     "total_labels": 630,
-    "total_preds": 632,
+    "total_preds": 631,
     "considered_empty": 0,
     "valid_preds": 630,
-    "invalid_preds": 2,
+    "invalid_preds": 1,
     "missing": 0,
-    "invalid_fraction": 0.00316,
-    "precision": 0.99684,
+    "invalid_fraction": 0.00158,
+    "precision": 0.99842,
     "recall": 1.0,
-    "f1": 0.99842,
-    "non_empty": 632
+    "f1": 0.99921,
+    "non_empty": 631
   },
   "mpjpe": {
     "count": 630,
-    "mean": 0.056116,
+    "mean": 0.056111,
     "median": 0.051456,
-    "std": 0.018382,
-    "sem": 0.000733,
+    "std": 0.018368,
+    "sem": 0.000732,
     "min": 0.028965,
     "max": 0.14306,
     "recall-0.025": 0.0,
@@ -2146,18 +2146,18 @@ Results of the model in various experiments on different datasets.
     "recall-0.5": 1.0,
     "num_labels": 630,
     "ap-0.025": 0.0,
-    "ap-0.05": 0.223916,
-    "ap-0.1": 0.928562,
+    "ap-0.05": 0.223683,
+    "ap-0.1": 0.928267,
     "ap-0.15": 0.999816,
     "ap-0.25": 0.999816,
     "ap-0.5": 0.999816
   },
   "head": {
     "count": 598,
-    "mean": 0.040794,
+    "mean": 0.040764,
     "median": 0.039496,
-    "std": 0.01381,
-    "sem": 0.000565,
+    "std": 0.01374,
+    "sem": 0.000562,
     "min": 0.011364,
     "max": 0.102955,
     "recall-0.025": 0.125418,
@@ -2170,15 +2170,15 @@ Results of the model in various experiments on different datasets.
   },
   "shoulder_left": {
     "count": 630,
-    "mean": 0.062854,
+    "mean": 0.062839,
     "median": 0.060457,
-    "std": 0.01943,
-    "sem": 0.000775,
+    "std": 0.019404,
+    "sem": 0.000774,
     "min": 0.018922,
     "max": 0.132634,
     "recall-0.025": 0.003175,
     "recall-0.05": 0.273016,
-    "recall-0.1": 0.974603,
+    "recall-0.1": 0.97619,
     "recall-0.15": 1.0,
     "recall-0.25": 1.0,
     "recall-0.5": 1.0,
@@ -2202,10 +2202,10 @@ Results of the model in various experiments on different datasets.
   },
   "elbow_left": {
     "count": 630,
-    "mean": 0.052412,
+    "mean": 0.05237,
     "median": 0.049508,
-    "std": 0.020244,
-    "sem": 0.000807,
+    "std": 0.020206,
+    "sem": 0.000806,
     "min": 0.010131,
     "max": 0.140634,
     "recall-0.025": 0.05873,
@@ -2234,10 +2234,10 @@ Results of the model in various experiments on different datasets.
   },
   "wrist_left": {
     "count": 630,
-    "mean": 0.048082,
+    "mean": 0.048071,
     "median": 0.041989,
-    "std": 0.026744,
-    "sem": 0.001066,
+    "std": 0.026751,
+    "sem": 0.001067,
     "min": 0.007895,
     "max": 0.191578,
     "recall-0.025": 0.134921,
@@ -2250,7 +2250,7 @@ Results of the model in various experiments on different datasets.
   },
   "wrist_right": {
     "count": 625,
-    "mean": 0.05271,
+    "mean": 0.052705,
     "median": 0.047416,
     "std": 0.025887,
     "sem": 0.001036,
@@ -2266,10 +2266,10 @@ Results of the model in various experiments on different datasets.
   },
   "hip_left": {
     "count": 630,
-    "mean": 0.057316,
+    "mean": 0.057311,
     "median": 0.054171,
-    "std": 0.020591,
-    "sem": 0.000821,
+    "std": 0.020577,
+    "sem": 0.00082,
     "min": 0.014001,
     "max": 0.17071,
     "recall-0.025": 0.025397,
@@ -2282,9 +2282,9 @@ Results of the model in various experiments on different datasets.
   },
   "hip_right": {
     "count": 629,
-    "mean": 0.055242,
+    "mean": 0.055245,
     "median": 0.050996,
-    "std": 0.02309,
+    "std": 0.023089,
     "sem": 0.000921,
     "min": 0.004999,
     "max": 0.145424,
@@ -2298,9 +2298,9 @@ Results of the model in various experiments on different datasets.
   },
   "knee_left": {
     "count": 628,
-    "mean": 0.045694,
+    "mean": 0.045693,
     "median": 0.034743,
-    "std": 0.04608,
+    "std": 0.046075,
     "sem": 0.00184,
     "min": 0.003593,
     "max": 0.364064,
@@ -2314,14 +2314,14 @@ Results of the model in various experiments on different datasets.
   },
   "knee_right": {
     "count": 629,
-    "mean": 0.053707,
+    "mean": 0.053711,
     "median": 0.036065,
-    "std": 0.072302,
+    "std": 0.072301,
     "sem": 0.002885,
     "min": 0.002669,
     "max": 0.496679,
     "recall-0.025": 0.252782,
-    "recall-0.05": 0.73132,
+    "recall-0.05": 0.72973,
     "recall-0.1": 0.934817,
     "recall-0.15": 0.952305,
     "recall-0.25": 0.958665,
@@ -2330,10 +2330,10 @@ Results of the model in various experiments on different datasets.
   },
   "ankle_left": {
     "count": 619,
-    "mean": 0.065828,
+    "mean": 0.065864,
     "median": 0.050491,
-    "std": 0.072587,
-    "sem": 0.00292,
+    "std": 0.072779,
+    "sem": 0.002928,
     "min": 0.012793,
     "max": 0.493666,
     "recall-0.025": 0.035541,
@@ -2346,7 +2346,7 @@ Results of the model in various experiments on different datasets.
   },
   "ankle_right": {
     "count": 601,
-    "mean": 0.054042,
+    "mean": 0.054043,
     "median": 0.047411,
     "std": 0.042559,
     "sem": 0.001737,
@@ -2363,8 +2363,8 @@ Results of the model in various experiments on different datasets.
   "joint_recalls": {
     "num_labels": 8129,
     "recall-0.025": 0.09583,
-    "recall-0.05": 0.52749,
-    "recall-0.1": 0.95079,
+    "recall-0.05": 0.52737,
+    "recall-0.1": 0.95092,
     "recall-0.15": 0.98155,
     "recall-0.25": 0.98905,
     "recall-0.5": 0.99742
diff --git a/scripts/test_skelda_dataset.py b/scripts/test_skelda_dataset.py
index 0c743ba..55d8053 100644
--- a/scripts/test_skelda_dataset.py
+++ b/scripts/test_skelda_dataset.py
@@ -333,13 +333,22 @@ def main():
         "koarob": 0.91,
     }
     minscore = minscores.get(dataset_use, 0.95)
+    min_group_sizes = {
+        # If the number of cameras is high, and the views are not occluded, use a higher value
+        "panoptic": 1,
+        "shelf": 2,
+        "tsinghua": 2,
+    }
+    min_group_size = min_group_sizes.get(dataset_use, 1)
+    if dataset_use == "panoptic" and len(datasets["panoptic"]["cams"]) == 10:
+        min_group_size = 5
 
     print("\nRunning predictions ...")
     all_poses = []
     all_ids = []
     all_paths = []
     times = []
-    triangulator = spt.Triangulator(min_score=minscore)
+    triangulator = spt.Triangulator(min_score=minscore, min_group_size=min_group_size)
     old_scene = ""
     for label in tqdm.tqdm(labels):
         images_2d = []
diff --git a/spt/interface.cpp b/spt/interface.cpp
index 0ac6fe6..83036d4 100644
--- a/spt/interface.cpp
+++ b/spt/interface.cpp
@@ -4,9 +4,9 @@
 // =================================================================================================
 // =================================================================================================
 
-Triangulator::Triangulator(float min_score)
+Triangulator::Triangulator(float min_score, size_t min_group_size)
 {
-    this->triangulator = new TriangulatorInternal(min_score);
+    this->triangulator = new TriangulatorInternal(min_score, min_group_size);
 }
 
 // =================================================================================================
diff --git a/spt/interface.hpp b/spt/interface.hpp
index 447d3a8..1020e0c 100644
--- a/spt/interface.hpp
+++ b/spt/interface.hpp
@@ -20,9 +20,11 @@ public:
      *
      *
      * @param min_score Minimum score to consider a triangulated joint as valid.
+     * @param min_group_size Minimum number of camera pairs that need to see a person.
      */
     Triangulator(
-        float min_score = 0.95);
+        float min_score = 0.95,
+        size_t min_group_size = 1);
 
     /**
      * Calculate a triangulation.
diff --git a/spt/triangulator.cpp b/spt/triangulator.cpp
index e510a30..d6aa6cd 100644
--- a/spt/triangulator.cpp
+++ b/spt/triangulator.cpp
@@ -102,9 +102,10 @@ void CameraInternal::update_projection_matrix()
 // =================================================================================================
 // =================================================================================================
 
-TriangulatorInternal::TriangulatorInternal(float min_score)
+TriangulatorInternal::TriangulatorInternal(float min_score, size_t min_group_size)
 {
     this->min_score = min_score;
+    this->min_group_size = min_group_size;
 }
 
 // =================================================================================================
@@ -443,20 +444,13 @@ std::vector<std::vector<std::array<float, 4>>> TriangulatorInternal::triangulate
     stime = std::chrono::high_resolution_clock::now();
 
     // Drop low scoring poses
-    std::vector<size_t> drop_indices;
-    for (size_t i = 0; i < all_scored_poses.size(); ++i)
+    size_t num_poses = all_scored_poses.size();
+    for (size_t i = num_poses; i > 0; --i)
     {
-        if (all_scored_poses[i].second < min_score)
+        if (all_scored_poses[i - 1].second < min_score)
         {
-            drop_indices.push_back(i);
-        }
-    }
-    if (!drop_indices.empty())
-    {
-        for (size_t i = drop_indices.size(); i > 0; --i)
-        {
-            all_scored_poses.erase(all_scored_poses.begin() + drop_indices[i - 1]);
-            all_pairs.erase(all_pairs.begin() + drop_indices[i - 1]);
+            all_scored_poses.erase(all_scored_poses.begin() + i - 1);
+            all_pairs.erase(all_pairs.begin() + i - 1);
         }
     }
 
@@ -464,6 +458,16 @@ std::vector<std::vector<std::array<float, 4>>> TriangulatorInternal::triangulate
     std::vector<std::tuple<cv::Point3d, cv::Mat, std::vector<int>>> groups;
     groups = calc_grouping(all_pairs, all_scored_poses, min_score);
 
+    // Drop groups with too few matches
+    size_t num_groups = groups.size();
+    for (size_t i = num_groups; i > 0; --i)
+    {
+        if (std::get<2>(groups[i - 1]).size() < this->min_group_size)
+        {
+            groups.erase(groups.begin() + i - 1);
+        }
+    }
+
     elapsed = std::chrono::high_resolution_clock::now() - stime;
     grouping_time += elapsed.count();
     stime = std::chrono::high_resolution_clock::now();
diff --git a/spt/triangulator.hpp b/spt/triangulator.hpp
index 4818d31..f569da8 100644
--- a/spt/triangulator.hpp
+++ b/spt/triangulator.hpp
@@ -31,7 +31,7 @@ public:
 class TriangulatorInternal
 {
 public:
-    TriangulatorInternal(float min_score);
+    TriangulatorInternal(float min_score, size_t min_group_size);
 
     std::vector<std::vector<std::array<float, 4>>> triangulate_poses(
         const std::vector<std::vector<std::vector<std::array<float, 3>>>> &poses_2d,
@@ -44,6 +44,8 @@ public:
 
 private:
     float min_score;
+    float min_group_size;
+
     const std::vector<std::string> core_joints = {
         "shoulder_left",
         "shoulder_right",