ultralytics 8.0.125 NMS speed improvements (#3463)

Co-authored-by: Compunet <117437050+dme-compunet@users.noreply.github.com>

ultralytics/__init__.py

@@ -1,6 +1,6 @@
 # Ultralytics YOLO 🚀, AGPL-3.0 license
 
-__version__ = '8.0.124'
+__version__ = '8.0.125'
 
 from ultralytics.hub import start
 from ultralytics.vit.rtdetr import RTDETR

ultralytics/nn/autobackend.py

@@ -340,7 +340,7 @@ class AutoBackend(nn.Module):
         elif self.coreml:  # CoreML
             im = im[0].cpu().numpy()
             im_pil = Image.fromarray((im * 255).astype('uint8'))
-            # im = im.resize((192, 320), Image.ANTIALIAS)
+            # im = im.resize((192, 320), Image.BILINEAR)
             y = self.model.predict({'image': im_pil})  # coordinates are xywh normalized
             if 'confidence' in y:
                 box = xywh2xyxy(y['coordinates'] * [[w, h, w, h]])  # xyxy pixels

ultralytics/yolo/engine/exporter.py

@@ -447,7 +447,7 @@ class Exporter:
                 check_requirements('nvidia-tensorrt', cmds='-U --index-url https://pypi.ngc.nvidia.com')
             import tensorrt as trt  # noqa
 
-        check_version(trt.__version__, '7.0.0', hard=True)  # require tensorrt>=8.0.0
+        check_version(trt.__version__, '7.0.0', hard=True)  # require tensorrt>=7.0.0
         self.args.simplify = True
         f_onnx, _ = self.export_onnx()
 

ultralytics/yolo/utils/ops.py

@@ -200,8 +200,7 @@ def non_max_suppression(
     multi_label &= nc > 1  # multiple labels per box (adds 0.5ms/img)
     merge = False  # use merge-NMS
 
-    prediction = prediction.clone()  # don't modify original
-    prediction = prediction.transpose(-1, -2)  # to (batch, boxes, items)
+    prediction = prediction.transpose(-1, -2)  # shape(1,84,6300) to shape(1,6300,84)
     prediction[..., :4] = xywh2xyxy(prediction[..., :4])  # xywh to xyxy
 
     t = time.time()
@@ -245,7 +244,6 @@ def non_max_suppression(
         n = x.shape[0]  # number of boxes
         if not n:  # no boxes
             continue
-
         if n > max_nms:  # excess boxes
             x = x[x[:, 4].argsort(descending=True)[:max_nms]]  # sort by confidence and remove excess boxes