dynamic resolution

vitookit/datasets/ffcv_transform.py

@@ -37,43 +37,85 @@ IMAGENET_STD = np.array([0.229, 0.224, 0.225]) * 255
 
 @gin.configurable
 class DynamicResolution:
-    def __init__(self, start_ramp=gin.REQUIRED, end_ramp=gin.REQUIRED,  min_res=160, max_res=224, step=32):        
+    def __init__(self, start_ramp=gin.REQUIRED, end_ramp=gin.REQUIRED,  
+                    res=(224,224,0),lower_scale=(0.2,0.2,0), upper_scale=(1,1,0)):
         self.start_ramp = start_ramp
         self.end_ramp = end_ramp
-        self.min_res = min_res
-        self.max_res = max_res
-        self.step = step
-        
-        assert min_res <= max_res
+        self.resolution = res
+        self.lower_scale = lower_scale
+        self.upper_scale = upper_scale
         
         
     def get_resolution(self, epoch):    
+        
+        lv,rv, step = self.resolution
+        
+        if step==0:
+            assert lv==rv
+            return rv
+        
+        if epoch <= self.start_ramp:
+            return lv
+
+        if epoch >= self.end_ramp:
+            return rv
+        
+        interp = np.interp([epoch], [self.start_ramp, self.end_ramp], [lv, rv])
+        num_steps = int(np.round((interp[0]-lv) / step))
+        return num_steps*step + lv
+    
+    def get_lower_scale(self, epoch):    
+        
+        lv,rv, step = self.lower_scale
+        
+        if step==0:
+            assert lv==rv
+            return rv
+        
+        if epoch <= self.start_ramp:
+            return lv
+
+        if epoch >= self.end_ramp:
+            return rv
+
+        
+        interp = np.interp([epoch], [self.start_ramp, self.end_ramp], [lv, rv])
+        num_steps = int(np.round((interp[0]-lv) / step))
+        return num_steps*step + lv
+    
+    def get_upper_scale(self, epoch):    
+        
+        lv,rv, step = self.upper_scale
+        
+        if step==0:
+            assert lv==rv
+            return rv
+        
         if epoch <= self.start_ramp:
-            return self.min_res
+            return lv
 
         if epoch >= self.end_ramp:
-            return self.max_res
+            return rv
 
-        # otherwise, linearly interpolate to the nearest multiple of 32
-        interp = np.interp([epoch], [self.start_ramp, self.end_ramp], [self.min_res, self.max_res])
-        final_res = int(np.round(interp[0] / self.step)) * self.step
-        return final_res
+        interp = np.interp([epoch], [self.start_ramp, self.end_ramp], [lv, rv])
+        num_steps = int(np.round((interp[0]-lv) / step))
+        return num_steps*step + lv
     
     def __call__(self, loader, epoch,is_ffcv=False):
         img_size = self.get_resolution(epoch)
-        print("resolution = %s" % str(img_size))
+        lower_scale = self.get_lower_scale(epoch)
+        upper_scale = self.get_upper_scale(epoch)
+        print(f"resolution = {img_size}, scale = ({lower_scale:.2f},{upper_scale:.2f}) ")
         if is_ffcv:
             pipeline=loader.pipeline_specs['image']
             if pipeline.decoder.output_size[0] != img_size:
-                k = img_size/self.max_res
-                pipeline.decoder.scale = (0.2*k,1)
+                pipeline.decoder.scale = (lower_scale,upper_scale)
                 pipeline.decoder.output_size = (img_size,img_size)
                 loader.generate_code()
         else:
             decoder = loader.dataset.transforms.transform.transforms[0]
-            k = img_size/self.max_res
             decoder.size=(img_size,img_size)
-            decoder.scale = (0.2*k,1)
+            decoder.scale = (lower_scale,upper_scale)
                 
                 
 @gin.configurable

vitookit/evaluation/eval_cls.py

@@ -29,7 +29,7 @@ from vitookit.datasets.transform import three_augmentation
 from vitookit.utils.helper import *
 from vitookit.utils import misc
 from vitookit.models.build_model import build_model
-from vitookit.datasets.build_dataset import build_dataset
+from vitookit.datasets.build_dataset import build_dataset, build_transform
 import wandb
 
 
@@ -104,6 +104,7 @@ def get_args_parser():
                         help='LR decay rate (default: 0.1)')
 
     # Augmentation parameters
+    
     parser.add_argument('--ThreeAugment', action='store_true', default=False) #3augment
     parser.add_argument('--src',action='store_true', default=False, 
                         help="Use Simple Random Crop (SRC) or Random Resized Crop (RRC). Use SRC when there is less risk of overfitting, such as on ImageNet-21k.")
@@ -276,7 +277,8 @@ def main(args):
     if args.ThreeAugment:
         transform = three_augmentation(args)
     else:
-        transform = None
+        transform = build_transform(is_train=True, args=args)
+    
     dataset_train, args.nb_classes = build_dataset(args=args, is_train=True, trnsfrm=transform)
     dataset_val, _ = build_dataset(is_train=False, args=args)