ultralytics 8.3.81__tar.gz → 8.3.83__tar.gz

{ultralytics-8.3.81/ultralytics.egg-info → ultralytics-8.3.83}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: ultralytics
-Version: 8.3.81
+Version: 8.3.83
 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-8.3.81 → ultralytics-8.3.83}/tests/test_exports.py

@@ -200,7 +200,6 @@ def test_export_paddle():
 
 
 @pytest.mark.slow
-@pytest.mark.skipif(IS_RASPBERRYPI, reason="MNN not supported on Raspberry Pi")
 def test_export_mnn():
     """Test YOLO exports to MNN format (WARNING: MNN test must precede NCNN test or CI error on Windows)."""
     file = YOLO(MODEL).export(format="mnn", imgsz=32)

{ultralytics-8.3.81 → ultralytics-8.3.83}/ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics 🚀 AGPL-3.0 License - https://ultralytics.com/license
 
-__version__ = "8.3.81"
+__version__ = "8.3.83"
 
 import os
 

{ultralytics-8.3.81 → ultralytics-8.3.83}/ultralytics/cfg/datasets/open-images-v7.yaml

@@ -633,6 +633,7 @@ download: |
   import warnings
 
   name = 'open-images-v7'
+  fo.config.dataset_zoo_dir = Path(SETTINGS["datasets_dir"]) / "fiftyone" / name
   fraction = 1.0  # fraction of full dataset to use
   LOGGER.warning('WARNING ⚠️ Open Images V7 dataset requires at least **561 GB of free space. Starting download...')
   for split in 'train', 'validation':  # 1743042 train, 41620 val images
@@ -642,7 +643,6 @@ download: |
       dataset = foz.load_zoo_dataset(name,
                                      split=split,
                                      label_types=['detections'],
-                                     dataset_dir=Path(SETTINGS['datasets_dir']) / 'fiftyone' / name,
                                      max_samples=round((1743042 if train else 41620) * fraction))
 
       # Define classes

{ultralytics-8.3.81 → ultralytics-8.3.83}/ultralytics/data/augment.py

@@ -1368,20 +1368,13 @@ class RandomHSV:
             img = labels["img"]
             dtype = img.dtype  # uint8
 
-            # Original implementation (bug) from ultralytics<=8.3.78
-            # r = np.random.uniform(-1, 1, 3) * [self.hgain, self.sgain, self.vgain] + 1  # random gains
-            # x = np.arange(0, 256, dtype=r.dtype)
-            # lut_hue = ((x * r[0]) % 180).astype(dtype)
-            # lut_sat = np.clip(x * r[1], 0, 255).astype(dtype)
-            # lut_val = np.clip(x * r[2], 0, 255).astype(dtype)
-
-            # Fixed implementation in https://github.com/ultralytics/ultralytics/pull/19311
-            r = np.random.uniform(-1, 1, 3) * (self.hgain, self.sgain, self.vgain) * (180, 255, 255)  # random gains
+            r = np.random.uniform(-1, 1, 3) * [self.hgain, self.sgain, self.vgain]  # random gains
             x = np.arange(0, 256, dtype=r.dtype)
-            lut_hue = ((x + r[0]) % 180).astype(dtype)
-            lut_sat = np.clip(x + r[1], 0, 255).astype(dtype)
-            lut_val = np.clip(x + r[2], 0, 255).astype(dtype)
-            lut_sat[0] = 0  # prevent pure white changing color
+            # lut_hue = ((x * (r[0] + 1)) % 180).astype(dtype)   # original hue implementation from ultralytics<=8.3.78
+            lut_hue = ((x + r[0] * 180) % 180).astype(dtype)
+            lut_sat = np.clip(x * (r[1] + 1), 0, 255).astype(dtype)
+            lut_val = np.clip(x * (r[2] + 1), 0, 255).astype(dtype)
+            lut_sat[0] = 0  # prevent pure white changing color, introduced in 8.3.79
 
             hue, sat, val = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))
             im_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val)))

{ultralytics-8.3.81 → ultralytics-8.3.83}/ultralytics/engine/exporter.py

@@ -62,6 +62,7 @@ import shutil
 import subprocess
 import time
 import warnings
+from contextlib import contextmanager
 from copy import deepcopy
 from datetime import datetime
 from pathlib import Path
@@ -183,6 +184,26 @@ def try_export(inner_func):
     return outer_func
 
 
+@contextmanager
+def arange_patch(args):
+    """
+    Workaround for ONNX torch.arange incompatibility with FP16.
+
+    https://github.com/pytorch/pytorch/issues/148041.
+    """
+    if args.dynamic and args.half and args.format == "onnx":
+        func = torch.arange
+
+        def arange(*args, dtype=None, **kwargs):
+            return func(*args, **kwargs).to(dtype)  # cast to dtype instead of passing dtype
+
+        torch.arange = arange  # patch
+        yield
+        torch.arange = func  # unpatch
+    else:
+        yield
+
+
 class Exporter:
     """
     A class for exporting a model.
@@ -263,7 +284,6 @@ class Exporter:
         if self.args.half and onnx and self.device.type == "cpu":
             LOGGER.warning("WARNING ⚠️ half=True only compatible with GPU export, i.e. use device=0")
             self.args.half = False
-            assert not self.args.dynamic, "half=True not compatible with dynamic=True, i.e. use only one."
         self.imgsz = check_imgsz(self.args.imgsz, stride=model.stride, min_dim=2)  # check image size
         if self.args.int8 and engine:
             self.args.dynamic = True  # enforce dynamic to export TensorRT INT8
@@ -515,7 +535,6 @@ class Exporter:
         output_names = ["output0", "output1"] if isinstance(self.model, SegmentationModel) else ["output0"]
         dynamic = self.args.dynamic
         if dynamic:
-            self.model.cpu()  # dynamic=True only compatible with cpu
             dynamic = {"images": {0: "batch", 2: "height", 3: "width"}}  # shape(1,3,640,640)
             if isinstance(self.model, SegmentationModel):
                 dynamic["output0"] = {0: "batch", 2: "anchors"}  # shape(1, 116, 8400)
@@ -533,17 +552,18 @@ class Exporter:
                 pass
             check_requirements("onnxslim>=0.1.46")  # Older versions has bug with OBB
 
-        torch.onnx.export(
-            NMSModel(self.model, self.args) if self.args.nms else self.model,
-            self.im.cpu() if dynamic else self.im,
-            f,
-            verbose=False,
-            opset_version=opset_version,
-            do_constant_folding=True,  # WARNING: DNN inference with torch>=1.12 may require do_constant_folding=False
-            input_names=["images"],
-            output_names=output_names,
-            dynamic_axes=dynamic or None,
-        )
+        with arange_patch(self.args):
+            torch.onnx.export(
+                NMSModel(self.model, self.args) if self.args.nms else self.model,
+                self.im,
+                f,
+                verbose=False,
+                opset_version=opset_version,
+                do_constant_folding=True,  # WARNING: DNN inference with torch>=1.12 may require do_constant_folding=False
+                input_names=["images"],
+                output_names=output_names,
+                dynamic_axes=dynamic or None,
+            )
 
         # Checks
         model_onnx = onnx.load(f)  # load onnx model

{ultralytics-8.3.81 → ultralytics-8.3.83/ultralytics.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: ultralytics
-Version: 8.3.81
+Version: 8.3.83
 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>