Updating yolo libs

models/common.py

@@ -354,6 +354,7 @@ class DetectMultiBackend(nn.Module):
             import onnxruntime
             providers = ['CUDAExecutionProvider', 'CPUExecutionProvider'] if cuda else ['CPUExecutionProvider']
             session = onnxruntime.InferenceSession(w, providers=providers)
+            output_names = [x.name for x in session.get_outputs()]
             meta = session.get_modelmeta().custom_metadata_map  # metadata
             if 'stride' in meta:
                 stride, names = int(meta['stride']), eval(meta['names'])
@@ -372,9 +373,7 @@ class DetectMultiBackend(nn.Module):
                 batch_size = batch_dim.get_length()
             executable_network = ie.compile_model(network, device_name="CPU")  # device_name="MYRIAD" for Intel NCS2
             output_layer = next(iter(executable_network.outputs))
-            meta = Path(w).with_suffix('.yaml')
+            stride, names = self._load_metadata(Path(w).with_suffix('.yaml'))  # load metadata
-            if meta.exists():
-                stride, names = self._load_metadata(meta)  # load metadata
         elif engine:  # TensorRT
             LOGGER.info(f'Loading {w} for TensorRT inference...')
             import tensorrt as trt  # https://developer.nvidia.com/nvidia-tensorrt-download
@@ -458,7 +457,7 @@ class DetectMultiBackend(nn.Module):
 
         self.__dict__.update(locals())  # assign all variables to self
 
-    def forward(self, im, augment=False, visualize=False, val=False):
+    def forward(self, im, augment=False, visualize=False):
         # YOLOv5 MultiBackend inference
         b, ch, h, w = im.shape  # batch, channel, height, width
         if self.fp16 and im.dtype != torch.float16:
@@ -466,17 +465,15 @@ class DetectMultiBackend(nn.Module):
 
         if self.pt:  # PyTorch
             y = self.model(im, augment=augment, visualize=visualize) if augment or visualize else self.model(im)
-            if isinstance(y, tuple):
-                y = y[0]
         elif self.jit:  # TorchScript
-            y = self.model(im)[0]
+            y = self.model(im)
         elif self.dnn:  # ONNX OpenCV DNN
             im = im.cpu().numpy()  # torch to numpy
             self.net.setInput(im)
             y = self.net.forward()
         elif self.onnx:  # ONNX Runtime
             im = im.cpu().numpy()  # torch to numpy
-            y = self.session.run([self.session.get_outputs()[0].name], {self.session.get_inputs()[0].name: im})[0]
+            y = self.session.run(self.output_names, {self.session.get_inputs()[0].name: im})
         elif self.xml:  # OpenVINO
             im = im.cpu().numpy()  # FP32
             y = self.executable_network([im])[self.output_layer]
@@ -523,9 +520,13 @@ class DetectMultiBackend(nn.Module):
                     y = (y.astype(np.float32) - zero_point) * scale  # re-scale
             y[..., :4] *= [w, h, w, h]  # xywh normalized to pixels
 
-        if isinstance(y, np.ndarray):
+        if isinstance(y, (list, tuple)):
-            y = torch.tensor(y, device=self.device)
+            return self.from_numpy(y[0]) if len(y) == 1 else [self.from_numpy(x) for x in y]
-        return (y, []) if val else y
+        else:
+            return self.from_numpy(y)
+
+    def from_numpy(self, x):
+        return torch.from_numpy(x).to(self.device) if isinstance(x, np.ndarray) else x
 
     def warmup(self, imgsz=(1, 3, 640, 640)):
         # Warmup model by running inference once
@@ -548,10 +549,12 @@ class DetectMultiBackend(nn.Module):
         return pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs
 
     @staticmethod
-    def _load_metadata(f='path/to/meta.yaml'):
+    def _load_metadata(f=Path('path/to/meta.yaml')):
         # Load metadata from meta.yaml if it exists
-        d = yaml_load(f)
+        if f.exists():
-        return d['stride'], d['names']  # assign stride, names
+            d = yaml_load(f)
+            return d['stride'], d['names']  # assign stride, names
+        return None, None
 
 
 class AutoShape(nn.Module):

utils/benchmarks.py

@@ -92,10 +92,14 @@ def run(
     LOGGER.info('\n')
     parse_opt()
     notebook_init()  # print system info
-    c = ['Format', 'Size (MB)', 'mAP@0.5:0.95', 'Inference time (ms)'] if map else ['Format', 'Export', '', '']
+    c = ['Format', 'Size (MB)', 'mAP50-95', 'Inference time (ms)'] if map else ['Format', 'Export', '', '']
     py = pd.DataFrame(y, columns=c)
     LOGGER.info(f'\nBenchmarks complete ({time.time() - t:.2f}s)')
     LOGGER.info(str(py if map else py.iloc[:, :2]))
+    if hard_fail and isinstance(hard_fail, str):
+        metrics = py['mAP50-95'].array  # values to compare to floor
+        floor = eval(hard_fail)  # minimum metric floor to pass, i.e. = 0.29 mAP for YOLOv5n
+        assert all(x > floor for x in metrics if pd.notna(x)), f'HARD FAIL: mAP50-95 < floor {floor}'
     return py
 
 
@@ -141,7 +145,7 @@ def parse_opt():
     parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
     parser.add_argument('--test', action='store_true', help='test exports only')
     parser.add_argument('--pt-only', action='store_true', help='test PyTorch only')
-    parser.add_argument('--hard-fail', action='store_true', help='throw error on benchmark failure')
+    parser.add_argument('--hard-fail', nargs='?', const=True, default=False, help='Exception on error or < min metric')
     opt = parser.parse_args()
     opt.data = check_yaml(opt.data)  # check YAML
     print_args(vars(opt))

utils/dataloaders.py

@@ -187,7 +187,7 @@ class _RepeatSampler:
 
 class LoadImages:
     # YOLOv5 image/video dataloader, i.e. `python detect.py --source image.jpg/vid.mp4`
-    def __init__(self, path, img_size=640, stride=32, auto=True, transforms=None):
+    def __init__(self, path, img_size=640, stride=32, auto=True, transforms=None, vid_stride=1):
         files = []
         for p in sorted(path) if isinstance(path, (list, tuple)) else [path]:
             p = str(Path(p).resolve())
@@ -212,6 +212,7 @@ class LoadImages:
         self.mode = 'image'
         self.auto = auto
         self.transforms = transforms  # optional
+        self.vid_stride = vid_stride  # video frame-rate stride
         if any(videos):
             self._new_video(videos[0])  # new video
         else:
@@ -232,6 +233,7 @@ class LoadImages:
             # Read video
             self.mode = 'video'
             ret_val, im0 = self.cap.read()
+            self.cap.set(cv2.CAP_PROP_POS_FRAMES, self.vid_stride * (self.frame + 1))  # read at vid_stride
             while not ret_val:
                 self.count += 1
                 self.cap.release()
@@ -242,7 +244,7 @@ class LoadImages:
                 ret_val, im0 = self.cap.read()
 
             self.frame += 1
-            # im0 = self._cv2_rotate(im0)  # for use if cv2 auto rotation is False
+            # im0 = self._cv2_rotate(im0)  # for use if cv2 autorotation is False
             s = f'video {self.count + 1}/{self.nf} ({self.frame}/{self.frames}) {path}: '
 
         else:
@@ -265,7 +267,7 @@ class LoadImages:
         # Create a new video capture object
         self.frame = 0
         self.cap = cv2.VideoCapture(path)
-        self.frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT))
+        self.frames = int(self.cap.get(cv2.CAP_PROP_FRAME_COUNT) / self.vid_stride)
         self.orientation = int(self.cap.get(cv2.CAP_PROP_ORIENTATION_META))  # rotation degrees
         # self.cap.set(cv2.CAP_PROP_ORIENTATION_AUTO, 0)  # disable https://github.com/ultralytics/yolov5/issues/8493
 
@@ -285,11 +287,12 @@ class LoadImages:
 
 class LoadStreams:
     # YOLOv5 streamloader, i.e. `python detect.py --source 'rtsp://example.com/media.mp4'  # RTSP, RTMP, HTTP streams`
-    def __init__(self, sources='streams.txt', img_size=640, stride=32, auto=True, transforms=None):
+    def __init__(self, sources='streams.txt', img_size=640, stride=32, auto=True, transforms=None, vid_stride=1):
         torch.backends.cudnn.benchmark = True  # faster for fixed-size inference
         self.mode = 'stream'
         self.img_size = img_size
         self.stride = stride
+        self.vid_stride = vid_stride  # video frame-rate stride
         sources = Path(sources).read_text().rsplit() if Path(sources).is_file() else [sources]
         n = len(sources)
         self.sources = [clean_str(x) for x in sources]  # clean source names for later
@@ -329,11 +332,11 @@ class LoadStreams:
 
     def update(self, i, cap, stream):
         # Read stream `i` frames in daemon thread
-        n, f, read = 0, self.frames[i], 1  # frame number, frame array, inference every 'read' frame
+        n, f = 0, self.frames[i]  # frame number, frame array
         while cap.isOpened() and n < f:
             n += 1
             cap.grab()  # .read() = .grab() followed by .retrieve()
-            if n % read == 0:
+            if n % self.vid_stride == 0:
                 success, im = cap.retrieve()
                 if success:
                     self.imgs[i] = im

utils/general.py

@@ -813,6 +813,9 @@ def non_max_suppression(prediction,
          list of detections, on (n,6) tensor per image [xyxy, conf, cls]
     """
 
+    if isinstance(prediction, (list, tuple)):  # YOLOv5 model in validation model, output = (inference_out, loss_out)
+        prediction = prediction[0]  # select only inference output
+
     bs = prediction.shape[0]  # batch size
     nc = prediction.shape[2] - 5  # number of classes
     xc = prediction[..., 4] > conf_thres  # candidates

utils/loggers/__init__.py

@@ -233,7 +233,9 @@ class Loggers():
             self.wandb.finish_run()
 
         if self.clearml and not self.opt.evolve:
-            self.clearml.task.update_output_model(model_path=str(best if best.exists() else last), name='Best Model')
+            self.clearml.task.update_output_model(model_path=str(best if best.exists() else last),
+                                                  name='Best Model',
+                                                  auto_delete_file=False)
 
     def on_params_update(self, params: dict):
         # Update hyperparams or configs of the experiment