Update use_dataset tutorial to integrate Albumentations for data augmentation

- Replaced torchvision transforms with Albumentations for image augmentation.
- Renumbered sections for clarity and updated descriptions accordingly.
- Emphasized key points for using Albumentations with 🤗 Datasets.

Author: ternaus

docs/source/use_dataset.mdx

@@ -175,37 +175,19 @@ Most image models expect the image to be in the RGB mode. The Beans images are a
 >>> dataset = dataset.cast_column("image", Image(mode="RGB"))
 ```
 
-**3**. Now, you can apply some transforms to the image. Feel free to take a look at the [various transforms available](https://docs.pytorch.org/vision/stable/transforms.html#v2-api-reference-recommended) in torchvision and choose one you'd like to experiment with. This example applies a transform that randomly rotates the image:
-
-```py
->>> from torchvision.transforms import RandomRotation
-
->>> rotate = RandomRotation(degrees=(0, 90))
->>> def transforms(examples):
-...     examples["pixel_values"] = [rotate(image) for image in examples["image"]]
-...     return examples
-```
-
-**4**. Use the [`~Dataset.set_transform`] function to apply the transform on-the-fly. When you index into the image `pixel_values`, the transform is applied, and your image gets rotated.
-
-```py
->>> dataset.set_transform(transforms)
->>> dataset[0]["pixel_values"]
-```
-
-**5**. The dataset is now ready for training with your machine learning framework!
+**3**. Now let's apply data augmentations to your images. 🤗 Datasets works with any augmentation library, and in this example we'll use Albumentations.
 
 ### Using Albumentations
 
-[Albumentations](https://albumentations.ai) is another popular image augmentation library that provides a [rich set of transforms](https://albumentations.ai/docs/reference/supported-targets-by-transform/) including spatial-level transforms, pixel-level transforms, and mixing-level transforms. When running on CPU, which is typical for transformers pipelines, Albumentations is [faster than torchvision](https://albumentations.ai/docs/benchmarks/image-benchmarks/).
+[Albumentations](https://albumentations.ai) is a popular image augmentation library that provides a [rich set of transforms](https://albumentations.ai/docs/reference/supported-targets-by-transform/) including spatial-level transforms, pixel-level transforms, and mixing-level transforms. When running on CPU, which is typical for transformers pipelines, Albumentations is [faster than torchvision](https://albumentations.ai/docs/benchmarks/image-benchmarks/).
 
-**1**. Install Albumentations:
+Install Albumentations:
 
 ```bash
 pip install albumentations
 ```
 
-**2**. Create a typical augmentation pipeline with Albumentations:
+**4**. Create a typical augmentation pipeline with Albumentations:
 
 ```py
 >>> import albumentations as A
@@ -219,7 +201,7 @@ pip install albumentations
 ... ])
 ```
 
-**3**. Since 🤗 Datasets uses PIL images but Albumentations expects OpenCV format (numpy arrays), you need to convert between formats:
+**5**. Since 🤗 Datasets uses PIL images but Albumentations expects OpenCV format (numpy arrays), you need to convert between formats:
 
 ```py
 >>> def albumentations_transforms(examples):
@@ -240,14 +222,14 @@ pip install albumentations
 ...     return examples
 ```
 
-**4**. Apply the transform using [`~Dataset.set_transform`]:
+**6**. Apply the transform using [`~Dataset.set_transform`]:
 
 ```py
 >>> dataset.set_transform(albumentations_transforms)
 >>> dataset[0]["pixel_values"]
 ```
 
-The key differences when using Albumentations:
+**Key points when using Albumentations with 🤗 Datasets:**
 - Convert PIL images to numpy arrays before applying transforms
 - Albumentations returns a dictionary with the transformed image under the "image" key
 - Convert the result back to PIL format after transformation