supervisely 6.73.286__py3-none-any.whl → 6.73.287__py3-none-any.whl

supervisely/nn/benchmark/base_benchmark.py

@@ -620,6 +620,9 @@ class BaseBenchmark:
     def get_eval_result(self):
         if self._eval_results is None:
             self._eval_results = self.evaluator.get_eval_result()
+            if not self._eval_results.inference_info:
+                self._eval_results.inference_info["gt_project_id"] = self.gt_project_info.id
+                self._eval_results.inference_info["dt_project_id"] = self.dt_project_info.id
         return self._eval_results
 
     def get_diff_project_info(self):

supervisely/nn/benchmark/base_evaluator.py

@@ -16,7 +16,7 @@ class BaseEvalResult:
 
     def __init__(self, directory: Optional[str] = None):
         self.directory = directory
-        self.inference_info: Dict = None
+        self.inference_info: Dict = {}
         self.speedtest_info: Dict = None
         self.eval_data: Dict = None
         self.mp = None

supervisely/nn/benchmark/instance_segmentation/text_templates.py

@@ -14,7 +14,7 @@ definitions = SimpleNamespace(
 )
 
 docs_url = (
-    "https://docs.supervisely.com/neural-networks/model-evaluation-benchmark/instant-segmentation"
+    "https://docs.supervisely.com/neural-networks/model-evaluation-benchmark/instance-segmentation"
 )
 
 # <i class="zmdi zmdi-check-circle" style="color: #13ce66; margin-right: 5px"></i>
@@ -44,6 +44,7 @@ markdown_overview = """
 {}
 - **IoU threshold**: {}
 - **Optimal confidence threshold**: {} (calculated automatically), <a href="{}" target="_blank">learn more</a>.
+- **Averaging across IoU thresholds:** {}, <a href="{}" target="_blank">learn more</a>.
 
 Learn more about Model Benchmark, implementation details, and how to use the charts in our <a href="{}" target="_blank">Technical Report</a>.
 """
@@ -60,7 +61,7 @@ Here, we comprehensively assess the model's performance by presenting a broad se
 - **Calibration Score**: This score represents the consistency of predicted probabilities (or <abbr title="{}">confidence scores</abbr>) made by the model. We evaluate how well predicted probabilities align with actual outcomes. A well-calibrated model means that when it predicts an object with, say, 80% confidence, approximately 80% of those predictions should actually be correct.
 """
 
-markdown_AP_custom_description = """> * AP_custom - Average Precision with different IoU thresholds for each class, that was set in evaluation params by the user."""
+markdown_AP_custom_description = """> *AP_custom - Average Precision with different IoU thresholds for each class, that was set in evaluation params by the user."""
 
 markdown_iou_per_class = """### IoU Threshold per Class
 

supervisely/nn/benchmark/object_detection/metric_provider.py

@@ -90,13 +90,16 @@ class MetricProvider:
         self.iouThrs = params["iouThrs"]
         self.recThrs = params["recThrs"]
 
+        # Evaluation params
         eval_params = params.get("evaluation_params", {})
         self.iou_threshold = eval_params.get("iou_threshold", 0.5)
         self.iou_threshold_idx = np.where(np.isclose(self.iouThrs, self.iou_threshold))[0][0]
-
-        # IoU per class (optional)
         self.iou_threshold_per_class = eval_params.get("iou_threshold_per_class")
         self.iou_idx_per_class = params.get("iou_idx_per_class")  # {cat id: iou_idx}
+        if self.iou_threshold_per_class is not None:
+            # TODO: temporary solution
+            eval_params["average_across_iou_thresholds"] = False
+        self.average_across_iou_thresholds = eval_params.get("average_across_iou_thresholds", True)
 
     def calculate(self):
         self.m_full = _MetricProvider(
@@ -199,9 +202,10 @@ class MetricProvider:
         }
 
     def AP_per_class(self):
-        s = self.coco_precision[:, :, :, 0, 2]
+        s = self.coco_precision[:, :, :, 0, 2].copy()
         s[s == -1] = np.nan
         ap = np.nanmean(s, axis=(0, 1))
+        ap = np.nan_to_num(ap, nan=0)
         return ap
 
     def AP_custom_per_class(self):
@@ -212,6 +216,7 @@ class MetricProvider:
                 s[:, cat_id - 1] = self.coco_precision[iou_idx, :, cat_id - 1, 0, 2]
         s[s == -1] = np.nan
         ap = np.nanmean(s, axis=0)
+        ap = np.nan_to_num(ap, nan=0)
         return ap
 
     def AP_custom(self):
@@ -284,6 +289,14 @@ class _MetricProvider:
         self.fp_not_confused_matches = [m for m in self.fp_matches if not m["miss_cls"]]
         self.ious = np.array([m["iou"] for m in self.tp_matches])
 
+        # Evaluation params
+        self.params = params
+        self.iou_idx_per_class = np.array(
+            [params["iou_idx_per_class"][cat_id] for cat_id in self.cat_ids]
+        )[:, None]
+        eval_params = params.get("evaluation_params", {})
+        self.average_across_iou_thresholds = eval_params.get("average_across_iou_thresholds", True)
+        
     def _init_counts(self):
         cat_ids = self.cat_ids
         iouThrs = self.iouThrs
@@ -328,14 +341,22 @@ class _MetricProvider:
         self.true_positives = true_positives
         self.false_negatives = false_negatives
         self.false_positives = false_positives
-        self.TP_count = int(self.true_positives[:, 0].sum(0))
-        self.FP_count = int(self.false_positives[:, 0].sum(0))
-        self.FN_count = int(self.false_negatives[:, 0].sum(0))
+        self.TP_count = int(self._take_iou_thresholds(true_positives).sum())
+        self.FP_count = int(self._take_iou_thresholds(false_positives).sum())
+        self.FN_count = int(self._take_iou_thresholds(false_negatives).sum())
 
+    def _take_iou_thresholds(self, x):
+        return np.take_along_axis(x, self.iou_idx_per_class, axis=1)
+    
     def base_metrics(self):
-        tp = self.true_positives
-        fp = self.false_positives
-        fn = self.false_negatives
+        if self.average_across_iou_thresholds:
+            tp = self.true_positives
+            fp = self.false_positives
+            fn = self.false_negatives
+        else:
+            tp = self._take_iou_thresholds(self.true_positives)
+            fp = self._take_iou_thresholds(self.false_positives)
+            fn = self._take_iou_thresholds(self.false_negatives)
         confuse_count = len(self.confused_matches)
 
         mAP = self.coco_mAP
@@ -358,9 +379,14 @@ class _MetricProvider:
         }
 
     def per_class_metrics(self):
-        tp = self.true_positives.mean(1)
-        fp = self.false_positives.mean(1)
-        fn = self.false_negatives.mean(1)
+        if self.average_across_iou_thresholds:
+            tp = self.true_positives.mean(1)
+            fp = self.false_positives.mean(1)
+            fn = self.false_negatives.mean(1)
+        else:
+            tp = self._take_iou_thresholds(self.true_positives).flatten()
+            fp = self._take_iou_thresholds(self.false_positives).flatten()
+            fn = self._take_iou_thresholds(self.false_negatives).flatten()
         pr = tp / (tp + fp)
         rc = tp / (tp + fn)
         f1 = 2 * pr * rc / (pr + rc)

supervisely/nn/benchmark/object_detection/text_templates.py

@@ -44,6 +44,7 @@ markdown_overview = """
 {}
 - **IoU threshold**: {}
 - **Optimal confidence threshold**: {} (calculated automatically), <a href="{}" target="_blank">learn more</a>.
+- **Averaging across IoU thresholds:** {}, <a href="{}" target="_blank">learn more</a>.
 
 Learn more about Model Benchmark, implementation details, and how to use the charts in our <a href="{}" target="_blank">Technical Report</a>.
 """
@@ -65,7 +66,7 @@ Here, we comprehensively assess the model's performance by presenting a broad se
 - **Calibration Score**: This score represents the consistency of predicted probabilities (or <abbr title="{}">confidence scores</abbr>) made by the model. We evaluate how well predicted probabilities align with actual outcomes. A well-calibrated model means that when it predicts an object with, say, 80% confidence, approximately 80% of those predictions should actually be correct.
 """
 
-markdown_AP_custom_description = """> * AP_custom - Average Precision with different IoU thresholds for each class, that was set in evaluation params by the user."""
+markdown_AP_custom_description = """> *AP_custom - Average Precision with different IoU thresholds for each class, that was set in evaluation params by the user."""
 
 markdown_iou_per_class = """### IoU Threshold per Class
 

supervisely/nn/benchmark/object_detection/vis_metrics/confidence_score.py

@@ -72,13 +72,13 @@ class ConfidenceScore(DetectionVisMetric):
             self.eval_result.dfsp_down,
             x="scores",
             y=["precision", "recall", "f1"],
-            labels={"value": "Value", "variable": "Metric", "scores": "Confidence Score"},
+            labels={"value": "Metric", "variable": "Metric", "scores": "Confidence Score"},
             width=None,
             height=500,
             color_discrete_map=color_map,
         )
         fig.update_traces(
-            hovertemplate="Confidence Score: %{x:.2f}<br>Value: %{y:.2f}<extra></extra>"
+            hovertemplate="Confidence score: %{x:.2f}<br>Metric: %{y:.2f}<extra></extra>"
         )
         fig.update_layout(yaxis=dict(range=[0, 1]), xaxis=dict(range=[0, 1], tick0=0, dtick=0.1))
 

supervisely/nn/benchmark/object_detection/vis_metrics/f1_score_at_different_iou.py

@@ -39,7 +39,7 @@ class F1ScoreAtDifferentIOU(DetectionVisMetric):
             np.concatenate([self.eval_result.dfsp_down["scores"].values[:, None], f1s_down.T], 1),
             columns=["scores"] + iou_names,
         )
-        labels = {"value": "Value", "variable": "IoU threshold", "scores": "Confidence Score"}
+        labels = {"value": "F1-score", "variable": "IoU threshold", "scores": "Confidence Score"}
 
         fig = px.line(
             df,
@@ -51,19 +51,19 @@ class F1ScoreAtDifferentIOU(DetectionVisMetric):
             height=500,
         )
         fig.update_traces(
-            hovertemplate="Confidence Score: %{x:.2f}<br>Value: %{y:.2f}<extra></extra>"
+            hovertemplate="Confidence Score: %{x:.2f}<br>F1-score: %{y:.2f}<extra></extra>"
         )
         fig.update_layout(yaxis=dict(range=[0, 1]), xaxis=dict(range=[0, 1], tick0=0, dtick=0.1))
 
         # add annotations for maximum F1-Score for each IoU threshold
         for i, iou in enumerate(iou_names):
-            argmax_f1 = f1s[i].argmax()
+            argmax_f1 = np.nanargmax(f1s[i])
             max_f1 = f1s[i][argmax_f1]
             score = self.eval_result.mp.m_full.score_profile["scores"][argmax_f1]
             fig.add_annotation(
                 x=score,
                 y=max_f1,
-                text=f"Best score: {score:.2f}",
+                text=f"Best conf: {score:.2f}",
                 showarrow=True,
                 arrowhead=1,
                 arrowcolor="black",

supervisely/nn/benchmark/object_detection/vis_metrics/outcome_counts_per_class.py

@@ -86,11 +86,10 @@ class PerClassOutcomeCounts(DetectionVisMetric):
         import plotly.express as px  # pylint: disable=import-error
 
         # Per-class Counts
-        iou_thres = 0
-
-        tp = self.eval_result.mp.true_positives[:, iou_thres]
-        fp = self.eval_result.mp.false_positives[:, iou_thres]
-        fn = self.eval_result.mp.false_negatives[:, iou_thres]
+        mp = self.eval_result.mp
+        tp = mp.m._take_iou_thresholds(mp.true_positives).flatten()
+        fp = mp.m._take_iou_thresholds(mp.false_positives).flatten()
+        fn = mp.m._take_iou_thresholds(mp.false_negatives).flatten()
 
         # normalize
         support = tp + fn

supervisely/nn/benchmark/object_detection/vis_metrics/overview.py

@@ -20,7 +20,7 @@ class Overview(DetectionVisMetric):
         url = self.eval_result.inference_info.get("checkpoint_url")
         link_text = self.eval_result.inference_info.get("custom_checkpoint_path")
         if link_text is None:
-            link_text = url
+            link_text = url or ""
         link_text = link_text.replace("_", "\_")
 
         model_name = self.eval_result.inference_info.get("model_name") or "Custom"
@@ -31,6 +31,7 @@ class Overview(DetectionVisMetric):
 
         # link to scroll to the optimal confidence section
         opt_conf_url = self.vis_texts.docs_url + "#f1-optimal-confidence-threshold"
+        average_url = self.vis_texts.docs_url + "#averaging-iou-thresholds"
 
         iou_threshold = self.eval_result.mp.iou_threshold
         if self.eval_result.different_iou_thresholds_per_class:
@@ -52,6 +53,8 @@ class Overview(DetectionVisMetric):
             iou_threshold,
             round(self.eval_result.mp.f1_optimal_conf, 4),
             opt_conf_url,
+            self.eval_result.mp.average_across_iou_thresholds,
+            average_url,
             self.vis_texts.docs_url,
         ]
 

supervisely/nn/benchmark/object_detection/vis_metrics/precision_avg_per_class.py

@@ -32,8 +32,7 @@ class PerClassAvgPrecision(DetectionVisMetric):
         import plotly.express as px  # pylint: disable=import-error
 
         # AP per-class
-        ap_per_class = self.eval_result.mp.coco_precision[:, :, :, 0, 2].mean(axis=(0, 1))
-        ap_per_class[ap_per_class == -1] = 0  # -1 is a placeholder for no GT
+        ap_per_class = self.eval_result.mp.AP_per_class()
         labels = dict(r="Average Precision", theta="Class")
         fig = px.scatter_polar(
             r=ap_per_class,

{supervisely-6.73.286.dist-info → supervisely-6.73.287.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: supervisely
-Version: 6.73.286
+Version: 6.73.287
 Summary: Supervisely Python SDK.
 Home-page: https://github.com/supervisely/supervisely
 Author: Supervisely

{supervisely-6.73.286.dist-info → supervisely-6.73.287.dist-info}/RECORD

@@ -744,8 +744,8 @@ supervisely/nn/artifacts/utils.py,sha256=C4EaMi95MAwtK5TOnhK4sQ1BWvgwYBxXyRStkhY
 supervisely/nn/artifacts/yolov5.py,sha256=slh05EpQsxqgKwB9KMClshdBxPBN3ZWZ6S4B80ECEt4,1724
 supervisely/nn/artifacts/yolov8.py,sha256=sFd9kU7Gdowq6WH1S3NdlQeoL9jjQKmRYb51fG_wbDk,1446
 supervisely/nn/benchmark/__init__.py,sha256=7jDezvavJFtO9mDeB2TqW8N4sD8TsHQBPpA9RESleIQ,610
-supervisely/nn/benchmark/base_benchmark.py,sha256=Xnb0jL0voBPC-s_eVYSYbYv-xVfLYtQf1tHLnJ9ktq8,25713
-supervisely/nn/benchmark/base_evaluator.py,sha256=sc8gNn3myGA8sGnP6EIiTp24JPXUQ9Ou-8BmTf-Dt7w,5248
+supervisely/nn/benchmark/base_benchmark.py,sha256=2buF7mD58igPMBPiEAJqfWHRO-JdPPzzOVlbR7-nvoA,25956
+supervisely/nn/benchmark/base_evaluator.py,sha256=MJeZnMcWr_cbeJ2r0GJ4SWgjWX5w33Y3pYVR6kCIQMQ,5246
 supervisely/nn/benchmark/base_visualizer.py,sha256=7woiYmztDzYZlbhL1hTfJnIi26RFi4obF2VLA519uxQ,10092
 supervisely/nn/benchmark/cv_tasks.py,sha256=ShoAbuNzfMYj0Se-KOnl_-dJnrmvN6Aukxa0eq28bFw,239
 supervisely/nn/benchmark/comparison/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -779,33 +779,33 @@ supervisely/nn/benchmark/instance_segmentation/__init__.py,sha256=47DEQpj8HBSa-_
 supervisely/nn/benchmark/instance_segmentation/benchmark.py,sha256=lTDzgKGpfeF5o_a2nS56wiAsUQPH1eubk37b9CaB2KI,1171
 supervisely/nn/benchmark/instance_segmentation/evaluation_params.yaml,sha256=POzpiaxnxuwAPSNQOGgjoUPfsk6Lf5hb9GLHwltWY5Y,94
 supervisely/nn/benchmark/instance_segmentation/evaluator.py,sha256=mpCi8S6YNwlVvgcERQSHBOhC9PrSfQkQ55pPTcK6V9c,2811
-supervisely/nn/benchmark/instance_segmentation/text_templates.py,sha256=_ZIU_3-xlUGKTcbEthxB4Ngt12azdC7pxpgqHHw7M3I,25780
+supervisely/nn/benchmark/instance_segmentation/text_templates.py,sha256=l2I1PbenuBzcCPu_h2J5JE5gQsJCr6lNnoIzk5BEuwc,25868
 supervisely/nn/benchmark/instance_segmentation/visualizer.py,sha256=8NscOKy7JK4AG-Czu3SM0qJQXLDfKD9URdG1d4nz89E,564
 supervisely/nn/benchmark/object_detection/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 supervisely/nn/benchmark/object_detection/base_vis_metric.py,sha256=XjUnFCnCMYLrpjojIOwiRNaSsSLYpozTHWfwLkaCd5U,1612
 supervisely/nn/benchmark/object_detection/benchmark.py,sha256=Wb4xlFXilIMVfsifNNQY25uE52NeEDLzQpnq8QPYq9U,1086
 supervisely/nn/benchmark/object_detection/evaluation_params.yaml,sha256=POzpiaxnxuwAPSNQOGgjoUPfsk6Lf5hb9GLHwltWY5Y,94
 supervisely/nn/benchmark/object_detection/evaluator.py,sha256=EOQQbmwQqjjvbRu3tY24SRA7K8nyqshR92gUcP1lcrY,7371
-supervisely/nn/benchmark/object_detection/metric_provider.py,sha256=MLVRnSwMQ9lfrlgBt4ThIHTVKY-6zuuEWK5-yVsmaj0,21140
-supervisely/nn/benchmark/object_detection/text_templates.py,sha256=J5xUPCGY-QWxc5AEt_u9_2r5q0LBlIzsa007H0GgoeU,26026
+supervisely/nn/benchmark/object_detection/metric_provider.py,sha256=iV79hlyB-_wj-X25-JPjoXfwQWNxGrMVweha3JZA46M,22557
+supervisely/nn/benchmark/object_detection/text_templates.py,sha256=ZjkcP91dj98_1xqxKSy5TGrU08puXaLhpjNC-c_41A0,26113
 supervisely/nn/benchmark/object_detection/visualizer.py,sha256=NpLKVW5fo6N0kYzgLsfY66wvCv38G3k-SNm4HImXt6g,32366
 supervisely/nn/benchmark/object_detection/vis_metrics/__init__.py,sha256=AXCLHEySEdR-B-5sfDoWBmmOLBVlyW2U_xr8Ta42sQI,2096
 supervisely/nn/benchmark/object_detection/vis_metrics/confidence_distribution.py,sha256=OlwkPgzEQ-RegcLZHVUVOL0n6I_2iayPVpAIie4y2O8,3615
-supervisely/nn/benchmark/object_detection/vis_metrics/confidence_score.py,sha256=r_saaZI4WB7C7ykNb1obmf8kEOkphLA4pInDoS6dXXU,4005
+supervisely/nn/benchmark/object_detection/vis_metrics/confidence_score.py,sha256=kuOhQDNwAsBbtjuMU_7GajVzu6j6n3xDJTv_hNKX6o8,4007
 supervisely/nn/benchmark/object_detection/vis_metrics/confusion_matrix.py,sha256=2PJUt0-njRpzN7XBGjkSt9kkh5tDPuv_Sne-2v8DWHc,3731
 supervisely/nn/benchmark/object_detection/vis_metrics/explore_predictions.py,sha256=wIYfq3izM2XNJHr56h3j5XhuU8W8Y3wO_RKAwxntQs4,4855
-supervisely/nn/benchmark/object_detection/vis_metrics/f1_score_at_different_iou.py,sha256=6y2Kx-R_t4SdJkdWNyZQ6TGjCC-u6KhXb4cCno4GuTk,2882
+supervisely/nn/benchmark/object_detection/vis_metrics/f1_score_at_different_iou.py,sha256=Aewzu2QhxZoPT_k0QJt_G11B8M2DXLCGRjE0MlVYNko,2892
 supervisely/nn/benchmark/object_detection/vis_metrics/frequently_confused.py,sha256=7rObk7WNsfwK7xBWl3aOxcn0uD48njEc04fQIPHc3_4,4678
 supervisely/nn/benchmark/object_detection/vis_metrics/iou_distribution.py,sha256=lv4Bk8W4X8ZhvQKyMXI46d240PNlMFx1hdji_aoTS50,3601
 supervisely/nn/benchmark/object_detection/vis_metrics/key_metrics.py,sha256=byucJuHYWSXIZU8U1Dc44QDpG3lTlhoNdUfD1b-uriw,4721
 supervisely/nn/benchmark/object_detection/vis_metrics/model_predictions.py,sha256=gsGDsesiwOcqeFvHr33b4PSJNw6MoA5brO-qRydRtsA,5944
 supervisely/nn/benchmark/object_detection/vis_metrics/outcome_counts.py,sha256=HuTgisYmXCSUeF5WOahy-uaCdvRLsNzg28BDrZ-5hww,7161
-supervisely/nn/benchmark/object_detection/vis_metrics/outcome_counts_per_class.py,sha256=GBq0KlPka5z4cxHcKCe2eVOI_h3qlWUqGCyhYs6mjrk,6825
-supervisely/nn/benchmark/object_detection/vis_metrics/overview.py,sha256=M6E--Yd1ztP4VBjR6VDUVrj2hgs5mwJF-vhWIjgVGkw,6376
+supervisely/nn/benchmark/object_detection/vis_metrics/outcome_counts_per_class.py,sha256=BKsb1XGVsg6-aOI5G6NibxvdD5lVzkMjHisI8T85Sns,6853
+supervisely/nn/benchmark/object_detection/vis_metrics/overview.py,sha256=uBxHIz8t0ujS5MI-LASg1RrPx3YqeKKc7dWz3MozHIU,6546
 supervisely/nn/benchmark/object_detection/vis_metrics/pr_curve.py,sha256=EeZmyNlTVQLQ-0wIDGdvFmRkahJBBiOKSmWiAJ8Bfks,3478
 supervisely/nn/benchmark/object_detection/vis_metrics/pr_curve_by_class.py,sha256=Bl_buVvH8SVqwsc4DcHnojMOqpwTnRgXFt9yw_Y1BR0,1607
 supervisely/nn/benchmark/object_detection/vis_metrics/precision.py,sha256=cAgMrp13uulHfM8xnPDZyR6PqS8nck1Fo7YPpvHPCbw,2708
-supervisely/nn/benchmark/object_detection/vis_metrics/precision_avg_per_class.py,sha256=X-hyf7OP7hzQzI5Yb2yDU536hxYQLvxcnVyYa3x27XA,2076
+supervisely/nn/benchmark/object_detection/vis_metrics/precision_avg_per_class.py,sha256=IetoRTE9xZHHMZlPwUmQyTnaD7cQpnIQ3zJmUUr6fgY,1965
 supervisely/nn/benchmark/object_detection/vis_metrics/recall.py,sha256=AAxg3eJDjWIZEarOMZNcdIcYpVDFFDlbc5bwF4X3GIo,2579
 supervisely/nn/benchmark/object_detection/vis_metrics/recall_vs_precision.py,sha256=u-00HH54XzlhFuzc7U-mk2-IhUSEPYzv23BIIz8STvk,1984
 supervisely/nn/benchmark/object_detection/vis_metrics/reliability_diagram.py,sha256=_8ie3dPfwOyNCg-YhqO5jFW6kubCfQC2Obn9QSZFVeQ,3615
@@ -1071,9 +1071,9 @@ supervisely/worker_proto/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZ
 supervisely/worker_proto/worker_api_pb2.py,sha256=VQfi5JRBHs2pFCK1snec3JECgGnua3Xjqw_-b3aFxuM,59142
 supervisely/worker_proto/worker_api_pb2_grpc.py,sha256=3BwQXOaP9qpdi0Dt9EKG--Lm8KGN0C5AgmUfRv77_Jk,28940
 supervisely_lib/__init__.py,sha256=7-3QnN8Zf0wj8NCr2oJmqoQWMKKPKTECvjH9pd2S5vY,159
-supervisely-6.73.286.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-supervisely-6.73.286.dist-info/METADATA,sha256=ulxDJ50Pdv9rflRKXgqgB8fpw1pOMrd4xzC1uDdIseU,33573
-supervisely-6.73.286.dist-info/WHEEL,sha256=P9jw-gEje8ByB7_hXoICnHtVCrEwMQh-630tKvQWehc,91
-supervisely-6.73.286.dist-info/entry_points.txt,sha256=U96-5Hxrp2ApRjnCoUiUhWMqijqh8zLR03sEhWtAcms,102
-supervisely-6.73.286.dist-info/top_level.txt,sha256=kcFVwb7SXtfqZifrZaSE3owHExX4gcNYe7Q2uoby084,28
-supervisely-6.73.286.dist-info/RECORD,,
+supervisely-6.73.287.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+supervisely-6.73.287.dist-info/METADATA,sha256=9oifX7yDdPoCm5GaVVjDyURQrPfVIjGIT27n3C10lf8,33573
+supervisely-6.73.287.dist-info/WHEEL,sha256=P9jw-gEje8ByB7_hXoICnHtVCrEwMQh-630tKvQWehc,91
+supervisely-6.73.287.dist-info/entry_points.txt,sha256=U96-5Hxrp2ApRjnCoUiUhWMqijqh8zLR03sEhWtAcms,102
+supervisely-6.73.287.dist-info/top_level.txt,sha256=kcFVwb7SXtfqZifrZaSE3owHExX4gcNYe7Q2uoby084,28
+supervisely-6.73.287.dist-info/RECORD,,