1 files changed, 13 insertions, 2 deletions
diff --git a/candle-transformers/src/models/vgg.rs b/candle-transformers/src/models/vgg.rs
index 010643c8..57f9ae67 100644
--- a/candle-transformers/src/models/vgg.rs
+++ b/candle-transformers/src/models/vgg.rs
@@ -1,7 +1,18 @@
 //! VGG-16 model implementation.
 //!
-//! See Very Deep Convolutional Networks for Large-Scale Image Recognition
-//! <https://arxiv.org/abs/1409.1556>
+//! VGG-16 is a convolutional neural network architecture. It consists of 13
+//! convolutional layers followed by 3 fully connected layers.
+//!
+//! Key characteristics:
+//! - Conv layers with 3x3 filters
+//! - Max pooling after every 2-3 conv layers
+//! - Three fully connected layers of 4096, 4096, 1000 units
+//! - ReLU activation and dropout
+//!
+//! References:
+//! - [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://arxiv.org/abs/1409.1556)
+//!
+
 use candle::{ModuleT, Result, Tensor};
 use candle_nn::{FuncT, VarBuilder};