Merge pull request #609 from mikel-brostrom/yolov5_v7_update

Yolov5 v7 update

reid_export.py

@@ -39,8 +39,6 @@
 from trackers.strong_sort.deep.models import build_model
 from trackers.strong_sort.deep.reid_model_factory import get_model_name, load_pretrained_weights
 
-# remove duplicated stream handler to avoid duplicated logging
-logging.getLogger().removeHandler(logging.getLogger().handlers[0])
 
 def file_size(path):
     # Return file/dir size (MB)

requirements.txt

@@ -1,55 +1,47 @@
 # pip install -r requirements.txt
 
-# base ----------------------------------------
+# Base ----------------------------------------
+gitpython
+ipython  # interactive notebook
 matplotlib>=3.2.2
 numpy>=1.18.5
 opencv-python>=4.1.1
 Pillow>=7.1.2
+psutil  # system resources
 PyYAML>=5.3.1
 requests>=2.23.0
 scipy>=1.4.1
-torch>=1.7.0
+thop>=0.1.1  # FLOPs computation
+torch>=1.7.0  # see https://pytorch.org/get-started/locally (recommended)
 torchvision>=0.8.1
 tqdm>=4.64.0
-protobuf<=3.20.1
+# protobuf<=3.20.1  # https://github.com/ultralytics/yolov5/issues/8012
 
-# plotting ------------------------------------
+# Logging ---------------------------------------------------------------------
+tensorboard>=2.4.1
+# clearml>=1.2.0
+# comet
 
+# Plotting --------------------------------------------------------------------
 pandas>=1.1.4
 seaborn>=0.11.0
 
-# deep_sort -----------------------------------
+# StrongSORT ------------------------------------------------------------------
 easydict
 
-# torchreid -----------------------------------
-
-Cython
-h5py
-six
-tb-nightly
-future
-yacs
+# torchreid -------------------------------------------------------------------
 gdown
-flake8
-yapf
-isort==4.3.21
-imageio
 
-#bytetrack
+# ByteTrack -------------------------------------------------------------------
 lap
 git+https://github.com/samson-wang/cython_bbox.git@9badb346a9222c98f828ba45c63fe3b7f2790ea2
 
-# ocsort
+# OCSORT ----------------------------------------------------------------------
 filterpy
 
-# Tracking eval ----------------------------------
-GitPython
-
-# Export --------------------------------------
-#onnx>=1.9.0  # ONNX export
-#onnx-simplifier>=0.4.1  # ONNX simplifier
-#tensorflow>=2.4.1  # TFLite export (or tensorflow-cpu, tensorflow-aarch64)
-#openvino-dev
-#git+https://github.com/PINTO0309/openvino2tensorflow #OpenVINO to TFLite export
+# Export ----------------------------------------------------------------------
+# onnx>=1.9.0  # ONNX export
+# onnx-simplifier>=0.4.1  # ONNX simplifier
 # nvidia-pyindex  # TensorRT export
-# nvidia-tensorrt  # TensorRT export
+# nvidia-tensorrt  # TensorRT export
+# openvino-dev  # OpenVINO export

track.py

@@ -30,14 +30,14 @@
 import logging
 from yolov5.models.common import DetectMultiBackend
 from yolov5.utils.dataloaders import VID_FORMATS, LoadImages, LoadStreams
-from yolov5.utils.general import (LOGGER, check_img_size, non_max_suppression, scale_coords, check_requirements, cv2,
+from yolov5.utils.general import (LOGGER, check_img_size, non_max_suppression, scale_boxes, check_requirements, cv2,
                                   check_imshow, xyxy2xywh, increment_path, strip_optimizer, colorstr, print_args, check_file)
 from yolov5.utils.torch_utils import select_device, time_sync
 from yolov5.utils.plots import Annotator, colors, save_one_box
 from trackers.multi_tracker_zoo import create_tracker
 
 # remove duplicated stream handler to avoid duplicated logging
-logging.getLogger().removeHandler(logging.getLogger().handlers[0])
+#logging.getLogger().removeHandler(logging.getLogger().handlers[0])
 
 @torch.no_grad()
 def run(
@@ -70,6 +70,7 @@ def run(
         hide_class=False,  # hide IDs
         half=False,  # use FP16 half-precision inference
         dnn=False,  # use OpenCV DNN for ONNX inference
+        vid_stride=1,  # video frame-rate stride
 ):
 
     source = str(source)
@@ -100,8 +101,7 @@ def run(
     # Dataloader
     if webcam:
         show_vid = check_imshow()
-        cudnn.benchmark = True  # set True to speed up constant image size inference
-        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt)
+        dataset = LoadStreams(source, img_size=imgsz, stride=stride, auto=pt, vid_stride=vid_stride)
         nr_sources = len(dataset)
     else:
         dataset = LoadImages(source, img_size=imgsz, stride=stride, auto=pt)
@@ -168,15 +168,15 @@ def run(
             s += '%gx%g ' % im.shape[2:]  # print string
             imc = im0.copy() if save_crop else im0  # for save_crop
 
-            annotator = Annotator(im0, line_width=line_thickness, pil=not ascii)
+            annotator = Annotator(im0, line_width=line_thickness, example=str(names))
 
             if hasattr(tracker_list[i], 'tracker') and hasattr(tracker_list[i].tracker, 'camera_update'):
                 if prev_frames[i] is not None and curr_frames[i] is not None:  # camera motion compensation
                     tracker_list[i].tracker.camera_update(prev_frames[i], curr_frames[i])
 
             if det is not None and len(det):
                 # Rescale boxes from img_size to im0 size
-                det[:, :4] = scale_coords(im.shape[2:], det[:, :4], im0.shape).round()  # xyxy
+                det[:, :4] = scale_boxes(im.shape[2:], det[:, :4], im0.shape).round()  # xyxy
 
                 # Print results
                 for c in det[:, -1].unique():
@@ -290,6 +290,7 @@ def parse_opt():
     parser.add_argument('--hide-class', default=False, action='store_true', help='hide IDs')
     parser.add_argument('--half', action='store_true', help='use FP16 half-precision inference')
     parser.add_argument('--dnn', action='store_true', help='use OpenCV DNN for ONNX inference')
+    parser.add_argument('--vid-stride', type=int, default=1, help='video frame-rate stride')
     opt = parser.parse_args()
     opt.imgsz *= 2 if len(opt.imgsz) == 1 else 1  # expand
     print_args(vars(opt))

trackers/strong_sort/reid_multibackend.py

@@ -34,8 +34,8 @@ def __init__(self, weights='osnet_x0_25_msmt17.pt', device=torch.device('cpu'),
         super().__init__()
 
         w = weights[0] if isinstance(weights, list) else weights
-        self.pt, self.jit, self.onnx, self.xml, self.engine, self.coreml, \
-            self.saved_model, self.pb, self.tflite, self.edgetpu, self.tfjs = self.model_type(w)  # get backend
+        self.pt, self.jit, self.onnx, self.xml, self.engine, self.coreml, self.saved_model, \
+            self.pb, self.tflite, self.edgetpu, self.tfjs, self.paddle = self.model_type(w)  # get backend
         self.fp16 = fp16
         self.fp16 &= self.pt or self.jit or self.engine  # FP16
 
@@ -164,14 +164,11 @@ def __init__(self, weights='osnet_x0_25_msmt17.pt', device=torch.device('cpu'),
     def model_type(p='path/to/model.pt'):
         # Return model type from model path, i.e. path='path/to/model.onnx' -> type=onnx
         from export import export_formats
-        suffixes = list(export_formats().Suffix) + ['.xml']  # export suffixes
-        check_suffix(p, suffixes)  # checks
-        p = Path(p).name  # eliminate trailing separators
-        pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs, xml2 = (s in p for s in suffixes)
-        xml |= xml2  # *_openvino_model or *.xml
-        tflite &= not edgetpu  # *.tflite
-        return pt, jit, onnx, xml, engine, coreml, saved_model, pb, tflite, edgetpu, tfjs
-
+        sf = list(export_formats().Suffix)  # export suffixes
+        check_suffix(p, sf)  # checks
+        types = [s in Path(p).name for s in sf]
+        types[8] &= not types[9]  # tflite &= not edgetpu
+        return types
 
     def _preprocess(self, im_batch):