updates

utils/datasets.py

@@ -626,9 +626,8 @@ def letterbox(img, new_shape=(416, 416), color=(114, 114, 114), auto=True, scale
 def random_affine(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10, border=0):
     # torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))
     # https://medium.com/uruvideo/dataset-augmentation-with-random-homographies-a8f4b44830d4
+    # targets = [cls, xyxy]
 
-    if targets is None:  # targets = [cls, xyxy]
-        targets = []
     height = img.shape[0] + border * 2
     width = img.shape[1] + border * 2
 
@@ -637,6 +636,7 @@ def random_affine(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10,
     a = random.uniform(-degrees, degrees)
     # a += random.choice([-180, -90, 0, 90])  # add 90deg rotations to small rotations
     s = random.uniform(1 - scale, 1 + scale)
+    # s = 2 ** random.uniform(-scale, scale)
     R[:2] = cv2.getRotationMatrix2D(angle=a, center=(img.shape[1] / 2, img.shape[0] / 2), scale=s)
 
     # Translation

utils/torch_utils.py

@@ -12,10 +12,10 @@ import torch.nn.functional as F
 def init_seeds(seed=0):
     torch.manual_seed(seed)
 
-    # Remove randomness (may be slower on Tesla GPUs) # https://pytorch.org/docs/stable/notes/randomness.html
+    # Reduce randomness (may be slower on Tesla GPUs) # https://pytorch.org/docs/stable/notes/randomness.html
     if seed == 0:
-        cudnn.deterministic = True
-        cudnn.benchmark = False
+        cudnn.deterministic = False
+        cudnn.benchmark = True
 
 
 def select_device(device='', apex=False, batch_size=None):

utils/utils.py

@@ -103,7 +103,7 @@ def coco80_to_coco91_class():  # converts 80-index (val2014) to 91-index (paper)
 
 
 def xyxy2xywh(x):
-    # Transform box coordinates from [x1, y1, x2, y2] (where xy1=top-left, xy2=bottom-right) to [x, y, w, h] 
+    # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h] where xy1=top-left, xy2=bottom-right
     y = torch.zeros_like(x) if isinstance(x, torch.Tensor) else np.zeros_like(x)
     y[:, 0] = (x[:, 0] + x[:, 2]) / 2  # x center
     y[:, 1] = (x[:, 1] + x[:, 3]) / 2  # y center
@@ -113,7 +113,7 @@ def xyxy2xywh(x):
 
 
 def xywh2xyxy(x):
-    # Transform box coordinates from [x, y, w, h] to [x1, y1, x2, y2] (where xy1=top-left, xy2=bottom-right)
+    # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right
     y = torch.zeros_like(x) if isinstance(x, torch.Tensor) else np.zeros_like(x)
     y[:, 0] = x[:, 0] - x[:, 2] / 2  # top left x
     y[:, 1] = x[:, 1] - x[:, 3] / 2  # top left y
@@ -122,26 +122,6 @@ def xywh2xyxy(x):
     return y
 
 
-# def xywh2xyxy(box):
-#     # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2]
-#     if isinstance(box, torch.Tensor):
-#         x, y, w, h = box.t()
-#         return torch.stack((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).t()
-#     else:  # numpy
-#         x, y, w, h = box.T
-#         return np.stack((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).T
-#
-#
-# def xyxy2xywh(box):
-#     # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h]
-#     if isinstance(box, torch.Tensor):
-#         x1, y1, x2, y2 = box.t()
-#         return torch.stack(((x1 + x2) / 2, (y1 + y2) / 2, x2 - x1, y2 - y1)).t()
-#     else:  # numpy
-#         x1, y1, x2, y2 = box.T
-#         return np.stack(((x1 + x2) / 2, (y1 + y2) / 2, x2 - x1, y2 - y1)).T
-
-
 def scale_coords(img1_shape, coords, img0_shape, ratio_pad=None):
     # Rescale coords (xyxy) from img1_shape to img0_shape
     if ratio_pad is None:  # calculate from img0_shape
@@ -187,7 +167,7 @@ def ap_per_class(tp, conf, pred_cls, target_cls):
 
     # Create Precision-Recall curve and compute AP for each class
     pr_score = 0.1  # score to evaluate P and R https://github.com/ultralytics/yolov3/issues/898
-    s = [len(unique_classes), tp.shape[1]]  # number class, number iou thresholds (i.e. 10 for mAP0.5...0.95)
+    s = [unique_classes.shape[0], tp.shape[1]]  # number class, number iou thresholds (i.e. 10 for mAP0.5...0.95)
     ap, p, r = np.zeros(s), np.zeros(s), np.zeros(s)
     for ci, c in enumerate(unique_classes):
         i = pred_cls == c
@@ -601,7 +581,7 @@ def print_model_biases(model):
         pass
 
 
-def strip_optimizer(f='weights/last.pt'):  # from utils.utils import *; strip_optimizer()
+def strip_optimizer(f='weights/best.pt'):  # from utils.utils import *; strip_optimizer()
     # Strip optimizer from *.pt files for lighter files (reduced by 2/3 size)
     x = torch.load(f, map_location=torch.device('cpu'))
     x['optimizer'] = None
@@ -614,12 +594,11 @@ def create_backbone(f='weights/last.pt'):  # from utils.utils import *; create_b
     x['optimizer'] = None
     x['training_results'] = None
     x['epoch'] = -1
-    for p in x['model'].values():
-        try:
+    for p in x['model'].parameters():
         p.requires_grad = True
-        except:
-            pass
-    torch.save(x, 'weights/backbone.pt')
+    s = 'weights/backbone.pt'
+    print('%s saved as %s' % (f, s))
+    torch.save(x, s)
 
 
 def coco_class_count(path='../coco/labels/train2014/'):