Update README.md

README.md

@@ -10,7 +10,7 @@ tags:
 
 ## Model description
 
-The model's architecture is a [U-Net](https://arxiv.org/abs/1505.04597) with identical input and output dimensions. U-Net is a popular semantic segmentation architecture that progressively downsamples and upsamples its input image, adding skip connections between layers having the same resolution. The network takes two inputs, the noisy images and the variances of their noise components, which it encodes using [sinusoidal embeddings](https://arxiv.org/abs/1706.03762). More details in the corresponding Keras code example.
+The model uses a [U-Net](https://arxiv.org/abs/1505.04597) with identical input and output dimensions. It progressively downsamples and upsamples its input image, adding skip connections between layers having the same resolution. The architecture is a simplified version of the architecture of [DDPM](https://arxiv.org/abs/2006.11239). It consists of convolutional residual blocks and lacks attention layers. The network takes two inputs, the noisy images and the variances of their noise components, which it encodes using [sinusoidal embeddings](https://arxiv.org/abs/1706.03762). More details in the Keras code example.
 
 ## Intended uses & limitations
 
@@ -18,15 +18,13 @@ The model is intended for educational purposes, as a simple example of denoising
 
 ## Training and evaluation data
 
-The model is trained on the [Oxford Flowers 102](https://www.tensorflow.org/datasets/catalog/oxford_flowers102) dataset for generating images, which is a diverse natural dataset containing around 8,000 images of flowers. Since the official splits are imbalanced (most of the images are contained in the test splite), I created new splits (80% train, 20% validation) using the [Tensorflow Datasets slicing API](https://www.tensorflow.org/datasets/splits). Center crops were used for preprocessing.
+The model is trained on the [Oxford Flowers 102](https://www.tensorflow.org/datasets/catalog/oxford_flowers102) dataset for generating images, which is a diverse natural dataset containing around 8,000 images of flowers. Since the official splits are imbalanced (most of the images are contained in the test splite), new random splits were created (80% train, 20% validation) for training the model. Center crops were used for preprocessing.
 
 ## Training procedure
 
-The model is trained to denoise noisy images, and can generate images by iteratively denoising pure Gaussian noise. More details in the corresponding Keras code example.
+The model is trained to denoise noisy images, and can generate images by iteratively denoising pure Gaussian noise. More details in the Keras code example.
 
-### Training hyperparameters
-
-The following hyperparameters were used during training:
+## Training hyperparameters
 
 | Hyperparameters | Value |
 | :-- | :-- |