move inference augmentation to model.forward()

detect.py

@@ -84,9 +84,13 @@ def detect(save_img=False):
 
         # Inference
         t1 = torch_utils.time_synchronized()
-        pred = model(img)[0].float() if half else model(img)[0]
+        pred = model(img, augment=opt.augment)[0]
         t2 = torch_utils.time_synchronized()
 
+        # to float
+        if half:
+            pred = pred.float()
+
         # Apply NMS
         pred = non_max_suppression(pred, opt.conf_thres, opt.iou_thres,
                                    multi_label=False, classes=opt.classes, agnostic=opt.agnostic_nms)
@@ -173,6 +177,7 @@ if __name__ == '__main__':
     parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
     parser.add_argument('--classes', nargs='+', type=int, help='filter by class')
     parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
     opt = parser.parse_args()
     print(opt)
 

models.py

@@ -226,13 +226,21 @@ class Darknet(nn.Module):
         self.seen = np.array([0], dtype=np.int64)  # (int64) number of images seen during training
         self.info(verbose)  # print model description
 
-    def forward(self, x, verbose=False):
+    def forward(self, x, augment=False, verbose=False):
-        img_size = x.shape[-2:]
+        img_size = x.shape[-2:]  # height, width
         yolo_out, out = [], []
         if verbose:
             print('0', x.shape)
             str = ''
 
+        # Augment images (inference and test only)
+        if augment:  # https://github.com/ultralytics/yolov3/issues/931
+            nb = x.shape[0]  # batch size
+            x = torch.cat((x,
+                           torch_utils.scale_img(x.flip(3), 0.9),  # flip-lr and scale
+                           torch_utils.scale_img(x, 0.7),  # scale
+                           ), 0)
+
         for i, module in enumerate(self.module_list):
             name = module.__class__.__name__
             if name in ['WeightedFeatureFusion', 'FeatureConcat']:  # sum, concat
@@ -256,9 +264,16 @@ class Darknet(nn.Module):
         elif ONNX_EXPORT:  # export
             x = [torch.cat(x, 0) for x in zip(*yolo_out)]
             return x[0], torch.cat(x[1:3], 1)  # scores, boxes: 3780x80, 3780x4
-        else:  # test
+        else:  # inference or test
-            io, p = zip(*yolo_out)  # inference output, training output
+            x, p = zip(*yolo_out)  # inference output, training output
-            return torch.cat(io, 1), p
+            x = torch.cat(x, 1)  # cat yolo outputs
+            if augment:  # de-augment results
+                x = torch.split(x, nb, dim=0)
+                x[1][..., :4] /= 0.9  # scale
+                x[1][..., 0] = img_size[1] - x[1][..., 0]  # flip lr
+                x[2][..., :4] /= 0.7  # scale
+                x = torch.cat(x, 1)
+            return x, p
 
     def fuse(self):
         # Fuse Conv2d + BatchNorm2d layers throughout model

test.py

@@ -88,26 +88,11 @@ def test(cfg,
 
         # Disable gradients
         with torch.no_grad():
-            # Augment images
-            if augment:  # https://github.com/ultralytics/yolov3/issues/931
-                imgs = torch.cat((imgs,
-                                  torch_utils.scale_img(imgs.flip(3), 0.9),  # flip-lr and scale
-                                  torch_utils.scale_img(imgs, 0.7),  # scale
-                                  ), 0)
-
             # Run model
             t = torch_utils.time_synchronized()
-            inf_out, train_out = model(imgs)  # inference and training outputs
+            inf_out, train_out = model(imgs, augment=augment)  # inference and training outputs
             t0 += torch_utils.time_synchronized() - t
 
-            # De-augment results
-            if augment:
-                x = torch.split(inf_out, nb, dim=0)
-                x[1][..., :4] /= 0.9  # scale
-                x[1][..., 0] = width - x[1][..., 0]  # flip lr
-                x[2][..., :4] /= 0.7  # scale
-                inf_out = torch.cat(x, 1)
-
             # Compute loss
             if hasattr(model, 'hyp'):  # if model has loss hyperparameters
                 loss += compute_loss(train_out, targets, model)[1][:3]  # GIoU, obj, cls
@@ -250,7 +235,7 @@ if __name__ == '__main__':
     parser.add_argument('--task', default='test', help="'test', 'study', 'benchmark'")
     parser.add_argument('--device', default='', help='device id (i.e. 0 or 0,1) or cpu')
     parser.add_argument('--single-cls', action='store_true', help='train as single-class dataset')
-    parser.add_argument('--augment', action='store_true', help='augmented testing')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
     opt = parser.parse_args()
     opt.save_json = opt.save_json or any([x in opt.data for x in ['coco.data', 'coco2014.data', 'coco2017.data']])
     print(opt)