ultralytics-opencv-headless 8.4.9__tar.gz → 8.4.10__tar.gz

{ultralytics_opencv_headless-8.4.9/ultralytics_opencv_headless.egg-info → ultralytics_opencv_headless-8.4.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ultralytics-opencv-headless
-Version: 8.4.9
+Version: 8.4.10
 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_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/tests/test_cuda.py

@@ -12,7 +12,7 @@ from ultralytics import YOLO
 from ultralytics.cfg import TASK2DATA, TASK2MODEL, TASKS
 from ultralytics.utils import ASSETS, IS_JETSON, WEIGHTS_DIR
 from ultralytics.utils.autodevice import GPUInfo
-from ultralytics.utils.checks import check_amp
+from ultralytics.utils.checks import check_amp, check_tensorrt
 from ultralytics.utils.torch_utils import TORCH_1_13
 
 # Try to find idle devices if CUDA is available
@@ -91,6 +91,7 @@ def test_export_onnx_matrix(task, dynamic, int8, half, batch, simplify, nms):
 )
 def test_export_engine_matrix(task, dynamic, int8, half, batch):
     """Test YOLO model export to TensorRT format for various configurations and run inference."""
+    check_tensorrt()
     import tensorrt as trt
 
     is_trt10 = int(trt.__version__.split(".", 1)[0]) >= 10

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/tests/test_python.py

@@ -702,8 +702,7 @@ def test_yoloe(tmp_path):
     # Predict
     # text-prompts
     model = YOLO(WEIGHTS_DIR / "yoloe-11s-seg.pt")
-    names = ["person", "bus"]
-    model.set_classes(names, model.get_text_pe(names))
+    model.set_classes(["person", "bus"])
     model(SOURCE, conf=0.01)
 
     from ultralytics import YOLOE

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-__version__ = "8.4.9"
+__version__ = "8.4.10"
 
 import importlib
 import os

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/engine/exporter.py

@@ -102,6 +102,7 @@ from ultralytics.utils import (
     callbacks,
     colorstr,
     get_default_args,
+    is_jetson,
 )
 from ultralytics.utils.checks import (
     IS_PYTHON_3_10,
@@ -110,6 +111,7 @@ from ultralytics.utils.checks import (
     check_executorch_requirements,
     check_imgsz,
     check_requirements,
+    check_tensorrt,
     check_version,
     is_intel,
     is_sudo_available,
@@ -1002,12 +1004,15 @@ class Exporter:
         assert self.im.device.type != "cpu", "export running on CPU but must be on GPU, i.e. use 'device=0'"
         f_onnx = self.export_onnx()  # run before TRT import https://github.com/ultralytics/ultralytics/issues/7016
 
+        # Force re-install TensorRT on CUDA 13 ARM devices to 10.15.x versions for RT-DETR exports
+        # https://github.com/ultralytics/ultralytics/issues/22873
+        if is_jetson(jetpack=7):
+            check_tensorrt("10.15")
+
         try:
             import tensorrt as trt
         except ImportError:
-            if LINUX:
-                cuda_version = torch.version.cuda.split(".")[0]
-                check_requirements(f"tensorrt-cu{cuda_version}>7.0.0,!=10.1.0")
+            check_tensorrt()
             import tensorrt as trt
         check_version(trt.__version__, ">=7.0.0", hard=True)
         check_version(trt.__version__, "!=10.1.0", msg="https://github.com/ultralytics/ultralytics/pull/14239")
@@ -1395,7 +1400,7 @@ class Exporter:
         nms.confidenceThresholdInputFeatureName = "confidenceThreshold"
         nms.iouThreshold = self.args.iou
         nms.confidenceThreshold = self.args.conf
-        nms.pickTop.perClass = True
+        nms.pickTop.perClass = not self.args.agnostic_nms
         nms.stringClassLabels.vector.extend(names.values())
         nms_model = ct.models.MLModel(nms_spec)
 

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/hub/session.py

@@ -12,7 +12,7 @@ from urllib.parse import parse_qs, urlparse
 
 from ultralytics import __version__
 from ultralytics.hub.utils import HELP_MSG, HUB_WEB_ROOT, PREFIX
-from ultralytics.utils import IS_COLAB, LOGGER, SETTINGS, TQDM, checks, emojis
+from ultralytics.utils import IS_COLAB, LOGGER, SETTINGS, TQDM, checks
 from ultralytics.utils.errors import HUBModelError
 
 AGENT_NAME = f"python-{__version__}-colab" if IS_COLAB else f"python-{__version__}-local"
@@ -121,7 +121,7 @@ class HUBTrainingSession:
         """
         self.model = self.client.model(model_id)
         if not self.model.data:  # then model does not exist
-            raise ValueError(emojis("❌ The specified HUB model does not exist"))  # TODO: improve error handling
+            raise HUBModelError(f"❌ Model not found: '{model_id}'. Verify the model ID is correct.")
 
         self.model_url = f"{HUB_WEB_ROOT}/models/{self.model.id}"
         if self.model.is_trained():
@@ -167,10 +167,8 @@ class HUBTrainingSession:
 
         self.model.create_model(payload)
 
-        # Model could not be created
-        # TODO: improve error handling
         if not self.model.id:
-            return None
+            raise HUBModelError(f"❌ Failed to create model '{self.filename}' on Ultralytics HUB. Please try again.")
 
         self.model_url = f"{HUB_WEB_ROOT}/models/{self.model.id}"
 

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/models/sam/predict.py

@@ -2619,7 +2619,7 @@ class SAM3VideoSemanticPredictor(SAM3SemanticPredictor):
         if not isinstance(orig_imgs, list):  # input images are a torch.Tensor, not a list
             orig_imgs = ops.convert_torch2numpy_batch(orig_imgs)
 
-        names = []
+        names = self.model.names if self.model.names != "visual" else {}
         if len(curr_obj_ids) == 0:
             pred_masks, pred_boxes = None, torch.zeros((0, 7), device=self.device)
         else:
@@ -2638,6 +2638,8 @@ class SAM3VideoSemanticPredictor(SAM3SemanticPredictor):
             pred_boxes = torch.cat(
                 [pred_boxes, pred_ids[keep][:, None], pred_scores[keep][..., None], pred_cls[keep][..., None]], dim=-1
             )
+            if pred_boxes.shape[0]:
+                names = names or dict(enumerate(str(i) for i in range(pred_boxes[:, 6].int().max() + 1)))
             if pred_masks.shape[0] > 1:
                 tracker_scores = torch.tensor(
                     [
@@ -2657,7 +2659,6 @@ class SAM3VideoSemanticPredictor(SAM3SemanticPredictor):
                         background_value=0,
                     ).squeeze(1)
                 ) > 0
-                names = self.model.names or dict(enumerate(str(i) for i in range(pred_boxes[:, 6].int().max())))
 
         results = []
         for masks, boxes, orig_img, img_path in zip([pred_masks], [pred_boxes], orig_imgs, self.batch[0]):

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/models/yolo/model.py

@@ -426,5 +426,6 @@ class YOLOE(Model):
                 self.predictor = None  # reset predictor
         elif isinstance(self.predictor, yolo.yoloe.YOLOEVPDetectPredictor):
             self.predictor = None  # reset predictor if no visual prompts
+        self.overrides["agnostic_nms"] = True  # use agnostic nms for YOLOE default
 
         return super().predict(source, stream, **kwargs)

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/nn/autobackend.py

@@ -339,7 +339,7 @@ class AutoBackend(nn.Module):
                 batch = metadata["batch"]
                 dynamic = metadata.get("args", {}).get("dynamic", dynamic)
             # OpenVINO inference modes are 'LATENCY', 'THROUGHPUT' (not recommended), or 'CUMULATIVE_THROUGHPUT'
-            inference_mode = "CUMULATIVE_THROUGHPUT" if batch > 1 and dynamic else "LATENCY"
+            inference_mode = "CUMULATIVE_THROUGHPUT" if dynamic and batch > 1 else "LATENCY"
             ov_compiled_model = core.compile_model(
                 ov_model,
                 device_name=device_name,

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/utils/__init__.py

@@ -762,7 +762,7 @@ def is_jetson(jetpack=None) -> bool:
     if jetson and jetpack:
         try:
             content = open("/etc/nv_tegra_release").read()
-            version_map = {4: "R32", 5: "R35", 6: "R36"}  # JetPack to L4T major version mapping
+            version_map = {4: "R32", 5: "R35", 6: "R36", 7: "R38"}  # JetPack to L4T major version mapping
             return jetpack in version_map and version_map[jetpack] in content
         except Exception:
             return False

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/utils/checks.py

@@ -507,6 +507,17 @@ def check_executorch_requirements():
     check_requirements("numpy<=2.3.5")
 
 
+def check_tensorrt(min_version: str = "7.0.0"):
+    """Check and install TensorRT requirements including platform-specific dependencies.
+
+    Args:
+        min_version (str): Minimum supported TensorRT version (default: "7.0.0").
+    """
+    if LINUX:
+        cuda_version = torch.version.cuda.split(".")[0]
+        check_requirements(f"tensorrt-cu{cuda_version}>={min_version},!=10.1.0")
+
+
 def check_torchvision():
     """Check the installed versions of PyTorch and Torchvision to ensure they're compatible.
 

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/utils/instance.py

@@ -408,7 +408,7 @@ class Instances:
         good = self.bbox_areas > 0
         if not all(good):
             self._bboxes = self._bboxes[good]
-            if len(self.segments):
+            if self.segments is not None and len(self.segments):
                 self.segments = self.segments[good]
             if self.keypoints is not None:
                 self.keypoints = self.keypoints[good]
@@ -482,3 +482,16 @@ class Instances:
     def bboxes(self) -> np.ndarray:
         """Return bounding boxes."""
         return self._bboxes.bboxes
+
+    def __repr__(self) -> str:
+        """Return a string representation of the Instances object."""
+        # Map private to public names and include direct attributes
+        attr_map = {"_bboxes": "bboxes"}
+        parts = []
+        for key, value in self.__dict__.items():
+            name = attr_map.get(key, key)
+            if name == "bboxes":
+                value = self.bboxes  # Use the property
+            if value is not None:
+                parts.append(f"{name}={value!r}")
+        return "Instances({})".format("\n".join(parts))

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10}/ultralytics/utils/loss.py

@@ -1193,8 +1193,6 @@ class TVPDetectLoss:
 
     def loss(self, preds: dict[str, torch.Tensor], batch: dict[str, torch.Tensor]) -> tuple[torch.Tensor, torch.Tensor]:
         """Calculate the loss for text-visual prompt detection."""
-        assert self.ori_reg_max == self.vp_criterion.reg_max  # TODO: remove it
-
         if self.ori_nc == preds["scores"].shape[1]:
             loss = torch.zeros(3, device=self.vp_criterion.device, requires_grad=True)
             return loss, loss.detach()
@@ -1230,8 +1228,6 @@ class TVPSegmentLoss(TVPDetectLoss):
 
     def loss(self, preds: Any, batch: dict[str, torch.Tensor]) -> tuple[torch.Tensor, torch.Tensor]:
         """Calculate the loss for text-visual prompt detection."""
-        assert self.ori_reg_max == self.vp_criterion.reg_max  # TODO: remove it
-
         if self.ori_nc == preds["scores"].shape[1]:
             loss = torch.zeros(4, device=self.vp_criterion.device, requires_grad=True)
             return loss, loss.detach()

{ultralytics_opencv_headless-8.4.9 → ultralytics_opencv_headless-8.4.10/ultralytics_opencv_headless.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ultralytics-opencv-headless
-Version: 8.4.9
+Version: 8.4.10
 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>