small updates on readme and typos

colabs/intriguing_properties/README.md

@@ -1,7 +1,15 @@
 ## Intriguing Properties of Contrastive Losses
 
-This folder contains code for "Intriguing Properties of Contrastive Losses".
+This folder contains code for the paper titled [Intriguing Properties of Contrastive Losses](https://arxiv.org/abs/2011.02803). And here is the [link](https://contrastive-learning.github.io/intriguing/) to its website/blog.
+
+```
+@article{chen2021intriguing,
+  title={Intriguing Properties of Contrastive Losses},
+  author={Chen, Ting and Luo, Calvin and Li, Lala},
+  journal={Advances in Neural Information Processing Systems},
+  volume={34},
+  year={2021}
+}
+```
 
-Technical report: [arxiv](https://arxiv.org/abs/2011.02803).
 
-The slides: [Google slides](https://docs.google.com/presentation/d/1nnXuqFGED-8BemqNeatbd15IIsz2naRMoQsTuTzLYqw/edit?usp=sharing).

data_util.py

@@ -238,7 +238,7 @@ def center_crop(image, height, width, crop_proportion):
   image_height = shape[0]
   image_width = shape[1]
   crop_height, crop_width = _compute_crop_shape(
-      image_height, image_width, height / width, crop_proportion)
+      image_height, image_width, width / height, crop_proportion)
   offset_height = ((image_height - crop_height) + 1) // 2
   offset_width = ((image_width - crop_width) + 1) // 2
   image = tf.image.crop_to_bounding_box(

model.py

@@ -62,7 +62,7 @@ def model_fn(features, labels, mode, params=None):
         # Finetune just supervised (linear) head will not update BN stats.
         model_train_mode = False
       else:
-        # Pretrain or finetuen anything else will update BN stats.
+        # Pretrain or finetune anything else will update BN stats.
         model_train_mode = is_training
       hiddens = model(features, is_training=model_train_mode)
 

tf2/data_util.py

@@ -234,7 +234,7 @@ def center_crop(image, height, width, crop_proportion):
   image_height = shape[0]
   image_width = shape[1]
   crop_height, crop_width = _compute_crop_shape(
-      image_height, image_width, height / width, crop_proportion)
+      image_height, image_width, width / height, crop_proportion)
   offset_height = ((image_height - crop_height) + 1) // 2
   offset_width = ((image_width - crop_width) + 1) // 2
   image = tf.image.crop_to_bounding_box(