Import YOLOv5 dataloader (#94)

Co-authored-by: Glenn Jocher <glenn.jocher@ultralytics.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

ultralytics/yolo/utils/__init__.py

@@ -6,13 +6,15 @@ import sys
 import threading
 from pathlib import Path
 
+import cv2
 import IPython
+import pandas as pd
 
 # Constants
 FILE = Path(__file__).resolve()
 ROOT = FILE.parents[2]  # YOLO
 RANK = int(os.getenv('RANK', -1))
-DATASETS_DIR = ROOT.parent / 'datasets'  # YOLOv5 datasets directory
+DATASETS_DIR = Path(os.getenv('YOLOv5_DATASETS_DIR', ROOT.parent / 'datasets'))  # global datasets directory
 NUM_THREADS = min(8, max(1, os.cpu_count() - 1))  # number of YOLOv5 multiprocessing threads
 AUTOINSTALL = str(os.getenv('YOLOv5_AUTOINSTALL', True)).lower() == 'true'  # global auto-install mode
 FONT = 'Arial.ttf'  # https://ultralytics.com/assets/Arial.ttf
@@ -20,6 +22,14 @@ VERBOSE = str(os.getenv('YOLOv5_VERBOSE', True)).lower() == 'true'  # global ver
 TQDM_BAR_FORMAT = '{l_bar}{bar:10}{r_bar}'  # tqdm bar format
 LOGGING_NAME = 'yolov5'
 
+# Settings
+# torch.set_printoptions(linewidth=320, precision=5, profile='long')
+# np.set_printoptions(linewidth=320, formatter={'float_kind': '{:11.5g}'.format})  # format short g, %precision=5
+pd.options.display.max_columns = 10
+cv2.setNumThreads(0)  # prevent OpenCV from multithreading (incompatible with PyTorch DataLoader)
+os.environ['NUMEXPR_MAX_THREADS'] = str(NUM_THREADS)  # NumExpr max threads
+os.environ['OMP_NUM_THREADS'] = '1' if platform.system() == 'darwin' else str(NUM_THREADS)  # OpenMP (PyTorch and SciPy)
+
 
 def is_colab():
     # Is environment a Google Colab instance?

ultralytics/yolo/utils/configs/default.yaml

@@ -2,13 +2,13 @@
 # Default training settings and hyperparameters for medium-augmentation COCO training
 
 # Task and Mode
-task: "classify" # choices=['detect', 'segment', 'classify', 'init'] # init is a special case
-mode: "train" # choice=['train', 'val', 'infer']
+task: "classify"  # choices=['detect', 'segment', 'classify', 'init'] # init is a special case
+mode: "train"  # choice=['train', 'val', 'infer']
 
 # Train settings -------------------------------------------------------------------------------------------------------
 model: null  # i.e. yolov5s.pt, yolo.yaml
 data: null  # i.e. coco128.yaml
-epochs: 300
+epochs: 100
 batch_size: 16
 imgsz: 640
 nosave: False
@@ -42,10 +42,10 @@ noval: False
 save_json: False
 save_hybrid: False
 conf_thres: 0.001
-iou_thres: 0.6
+iou_thres: 0.7
 max_det: 300
 half: True
-dnn: False # use OpenCV DNN for ONNX inference
+dnn: False  # use OpenCV DNN for ONNX inference
 plots: True
 
 # Prediction settings:
@@ -56,9 +56,9 @@ save_conf: False
 save_crop: False
 hide_labels: False  # hide labels
 hide_conf: False
-vid_stride: 1 # video frame-rate stride
+vid_stride: 1  # video frame-rate stride
 line_thickness: 3  # bounding box thickness (pixels)
-update: False # Update all models
+update: False  # Update all models
 visualize: False
 augment: False
 agnostic_nms: False  # class-agnostic NMS
@@ -77,7 +77,7 @@ cls: 0.5  # cls loss gain (scale with pixels)
 dfl: 1.5  # dfl loss gain
 fl_gamma: 0.0  # focal loss gamma (efficientDet default gamma=1.5)
 label_smoothing: 0.0
-nbs: 64 # nominal batch size
+nbs: 64  # nominal batch size
 hsv_h: 0.015  # image HSV-Hue augmentation (fraction)
 hsv_s: 0.7  # image HSV-Saturation augmentation (fraction)
 hsv_v: 0.4  # image HSV-Value augmentation (fraction)
@@ -92,6 +92,9 @@ mosaic: 1.0  # image mosaic (probability)
 mixup: 0.0  # image mixup (probability)
 copy_paste: 0.0  # segment copy-paste (probability)
 
+# For debugging. Don't change
+v5loader: True
+
 # Hydra configs --------------------------------------------------------------------------------------------------------
 hydra:
   output_subdir: null  # disable hydra directory creation

ultralytics/yolo/utils/dist.py

@@ -47,7 +47,7 @@ def generate_ddp_command(world_size, trainer):
     if using_cli:
         file_name = generate_ddp_file(trainer)
     return [
-        sys.executable, "-m", "torch.distributed.launch", "--nproc_per_node", f"{world_size}", "--master_port",
+        sys.executable, "-m", "torch.distributed.run", "--nproc_per_node", f"{world_size}", "--master_port",
         f"{find_free_network_port()}", file_name] + sys.argv[1:]
 
 
@@ -55,7 +55,7 @@ def ddp_cleanup(command, trainer):
     # delete temp file if  created
     # TODO: this is a temp solution in case the file is deleted before DDP launching
     time.sleep(5)
-    tempfile_suffix = str(id(trainer)) + ".py"
+    tempfile_suffix = f"{id(trainer)}.py"
     if tempfile_suffix in "".join(command):
         for chunk in command:
             if tempfile_suffix in chunk: