fix preset (#5454)

tutorials/mobilenetv3_prod/Step6/paddlevision/transforms/transforms.py

@@ -17,8 +17,13 @@ from . import functional as F
 from .functional import InterpolationMode, _interpolation_modes_from_int
 
 __all__ = [
-    "Compose", "ToTensor", "Normalize", "Resize", "CenterCrop",
-    "RandomResizedCrop", "RandomHorizontalFlip"
+    "Compose",
+    "ToTensor",
+    "Normalize",
+    "Resize",
+    "CenterCrop",
+    "RandomResizedCrop",
+    "RandomHorizontalFlip",
 ]
 
 

tutorials/mobilenetv3_prod/Step6/presets.py

+import paddle
+
 from paddlevision.transforms import autoaugment, transforms
 
 
@@ -12,16 +14,11 @@ class ClassificationPresetTrain:
         trans = [transforms.RandomResizedCrop(crop_size)]
         if hflip_prob > 0:
             trans.append(transforms.RandomHorizontalFlip(hflip_prob))
-        #if auto_augment_policy is not None:
-        #    aa_policy = autoaugment.AutoAugmentPolicy(auto_augment_policy)
-        #    trans.append(autoaugment.AutoAugment(policy=aa_policy))
         trans.extend([
             transforms.ToTensor(),
             transforms.Normalize(
                 mean=mean, std=std),
         ])
-        #if random_erase_prob > 0:
-        #    trans.append(transforms.RandomErasing(p=random_erase_prob))
 
         self.transforms = transforms.Compose(trans)
 
@@ -35,12 +32,14 @@ class ClassificationPresetEval:
                  resize_size=256,
                  mean=(0.485, 0.456, 0.406),
                  std=(0.229, 0.224, 0.225)):
-
+        mean = tuple([m * 255 for m in mean])
+        std = tuple([s * 255 for s in std])
         self.transforms = transforms.Compose([
             transforms.Resize(resize_size),
             transforms.CenterCrop(crop_size),
-            transforms.ToTensor(),
-            transforms.Normalize(
+            # fix to support pt-quant
+            paddle.vision.transforms.Transpose((2, 0, 1)),
+            paddle.vision.transforms.Normalize(
                 mean=mean, std=std),
         ])
 

tutorials/mobilenetv3_prod/Step6/tools/predict.py

@@ -49,6 +49,7 @@ def main(args):
         img = Image.open(f).convert('RGB')
 
     img = eval_transforms(img)
+    img = paddle.to_tensor(img)
     img = img.expand([1] + img.shape)
 
     output = model(img).numpy()[0]