ultralytics 8.3.49__tar.gz → 8.3.50__tar.gz

{ultralytics-8.3.49/ultralytics.egg-info → ultralytics-8.3.50}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ultralytics
-Version: 8.3.49
+Version: 8.3.50
 Summary: Ultralytics YOLO 🚀 for SOTA object detection, multi-object tracking, instance segmentation, pose estimation and image classification.
 Author-email: Glenn Jocher <glenn.jocher@ultralytics.com>, Jing Qiu <jing.qiu@ultralytics.com>
 Maintainer-email: Ultralytics <hello@ultralytics.com>

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics YOLO 🚀, AGPL-3.0 license
 
-__version__ = "8.3.49"
+__version__ = "8.3.50"
 
 import os
 

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/data/dataset.py

@@ -218,8 +218,10 @@ class YOLODataset(BaseDataset):
         # NOTE: do NOT resample oriented boxes
         segment_resamples = 100 if self.use_obb else 1000
         if len(segments) > 0:
-            # list[np.array(1000, 2)] * num_samples
-            # (N, 1000, 2)
+            # make sure segments interpolate correctly if original length is greater than segment_resamples
+            max_len = max([len(s) for s in segments])
+            segment_resamples = (max_len + 1) if segment_resamples < max_len else segment_resamples
+            # list[np.array(segment_resamples, 2)] * num_samples
             segments = np.stack(resample_segments(segments, n=segment_resamples), axis=0)
         else:
             segments = np.zeros((0, segment_resamples, 2), dtype=np.float32)

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/engine/model.py

@@ -115,7 +115,7 @@ class Model(nn.Module):
         self.predictor = None  # reuse predictor
         self.model = None  # model object
         self.trainer = None  # trainer object
-        self.ckpt = None  # if loaded from *.pt
+        self.ckpt = {}  # if loaded from *.pt
         self.cfg = None  # if loaded from *.yaml
         self.ckpt_path = None
         self.overrides = {}  # overrides for trainer object
@@ -807,7 +807,7 @@ class Model(nn.Module):
         # Update model and cfg after training
         if RANK in {-1, 0}:
             ckpt = self.trainer.best if self.trainer.best.exists() else self.trainer.last
-            self.model, _ = attempt_load_one_weight(ckpt)
+            self.model, self.ckpt = attempt_load_one_weight(ckpt)
             self.overrides = self.model.args
             self.metrics = getattr(self.trainer.validator, "metrics", None)  # TODO: no metrics returned by DDP
         return self.metrics

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/engine/validator.py

@@ -120,7 +120,7 @@ class BaseValidator:
             self.args.plots &= trainer.stopper.possible_stop or (trainer.epoch == trainer.epochs - 1)
             model.eval()
         else:
-            if str(self.args.model).endswith(".yaml"):
+            if str(self.args.model).endswith(".yaml") and model is None:
                 LOGGER.warning("WARNING ⚠️ validating an untrained model YAML will result in 0 mAP.")
             callbacks.add_integration_callbacks(self)
             model = AutoBackend(

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/models/sam/modules/sam.py

@@ -685,11 +685,11 @@ class SAM2Model(torch.nn.Module):
                 if prev is None:
                     continue  # skip padding frames
                 # "maskmem_features" might have been offloaded to CPU in demo use cases,
-                # so we load it back to GPU (it's a no-op if it's already on GPU).
-                feats = prev["maskmem_features"].cuda(non_blocking=True)
+                # so we load it back to inference device (it's a no-op if it's already on device).
+                feats = prev["maskmem_features"].to(device=device, non_blocking=True)
                 to_cat_memory.append(feats.flatten(2).permute(2, 0, 1))
                 # Spatial positional encoding (it might have been offloaded to CPU in eval)
-                maskmem_enc = prev["maskmem_pos_enc"][-1].cuda()
+                maskmem_enc = prev["maskmem_pos_enc"][-1].to(device=device)
                 maskmem_enc = maskmem_enc.flatten(2).permute(2, 0, 1)
                 # Temporal positional encoding
                 maskmem_enc = maskmem_enc + self.maskmem_tpos_enc[self.num_maskmem - t_pos - 1]

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/models/yolo/obb/train.py

@@ -39,4 +39,6 @@ class OBBTrainer(yolo.detect.DetectionTrainer):
     def get_validator(self):
         """Return an instance of OBBValidator for validation of YOLO model."""
         self.loss_names = "box_loss", "cls_loss", "dfl_loss"
-        return yolo.obb.OBBValidator(self.test_loader, save_dir=self.save_dir, args=copy(self.args))
+        return yolo.obb.OBBValidator(
+            self.test_loader, save_dir=self.save_dir, args=copy(self.args), _callbacks=self.callbacks
+        )

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/utils/instance.py

@@ -7,7 +7,7 @@ from typing import List
 
 import numpy as np
 
-from .ops import ltwh2xywh, ltwh2xyxy, xywh2ltwh, xywh2xyxy, xyxy2ltwh, xyxy2xywh
+from .ops import ltwh2xywh, ltwh2xyxy, resample_segments, xywh2ltwh, xywh2xyxy, xyxy2ltwh, xyxy2xywh
 
 
 def _ntuple(n):
@@ -406,7 +406,20 @@ class Instances:
         normalized = instances_list[0].normalized
 
         cat_boxes = np.concatenate([ins.bboxes for ins in instances_list], axis=axis)
-        cat_segments = np.concatenate([b.segments for b in instances_list], axis=axis)
+        seg_len = [b.segments.shape[1] for b in instances_list]
+        if len(set(seg_len)) > 1:  # resample segments if there's different length
+            max_len = max(seg_len)
+            cat_segments = np.concatenate(
+                [
+                    resample_segments(list(b.segments), max_len)
+                    if len(b.segments)
+                    else np.zeros((0, max_len, 2), dtype=np.float32)  # re-generating empty segments
+                    for b in instances_list
+                ],
+                axis=axis,
+            )
+        else:
+            cat_segments = np.concatenate([b.segments for b in instances_list], axis=axis)
         cat_keypoints = np.concatenate([b.keypoints for b in instances_list], axis=axis) if use_keypoint else None
         return cls(cat_boxes, cat_segments, cat_keypoints, bbox_format, normalized)
 

{ultralytics-8.3.49 → ultralytics-8.3.50}/ultralytics/utils/ops.py

@@ -624,6 +624,8 @@ def resample_segments(segments, n=1000):
         segments (list): the resampled segments.
     """
     for i, s in enumerate(segments):
+        if len(s) == n:
+            continue
         s = np.concatenate((s, s[0:1, :]), axis=0)
         x = np.linspace(0, len(s) - 1, n - len(s) if len(s) < n else n)
         xp = np.arange(len(s))

{ultralytics-8.3.49 → ultralytics-8.3.50/ultralytics.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: ultralytics
-Version: 8.3.49
+Version: 8.3.50
 Summary: Ultralytics YOLO 🚀 for SOTA object detection, multi-object tracking, instance segmentation, pose estimation and image classification.
 Author-email: Glenn Jocher <glenn.jocher@ultralytics.com>, Jing Qiu <jing.qiu@ultralytics.com>
 Maintainer-email: Ultralytics <hello@ultralytics.com>