Update README.md

README.md

@@ -9,7 +9,8 @@ This Repository is an official implementation of "[Compact3D: Compressing Gaussi
 Compact3D is a method to reduce the storage memory requirements of [3D Gaussian Splatting](https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/) models. 3D Gaussian Splatting is a new technique for novel view synthesis where properties of 3D Gaussians (location, shape, color) are optimized to model a 3D scene. The method performs better than SOTA NeRF approaches, is extremely fast to train and can be rendered in real time during inference. However, since a typical scene requires millions of Gaussians to model it, the memory requirements can be an order of magnitude more than many NeRF approaches. Here, we reduce the size of the trained 3D Gaussian Splat models by 10-20x by vector quantizing the Gaussian parameters. An overview of our method is shown below. We perform K-Means quantization on the covariance and color parameters of all Gaussians and replace values of each with the corresponding entry in the codebook (i.e., the cluster center). This is done in conjuncion with the training of the parameter values as done in the non-quantized version of Gaussian splatting. We observe that the models can be compressed 20 times without a big drop in performance. 
 
 ## Updates
-**[16 February 2024]:** Trained models are now available [here](https://drive.google.com/drive/folders/14WVjkqmkhbJQ-IEM_Un0YTTiB2jKJqQD?usp=sharing). Models are available for MipNerf-360, Tanks and Temples and DeepBlending dataset for both CompGS-4k and CompGS-32k variants.
+**[16 February 2024]:** Trained models are now available [here](https://drive.google.com/drive/folders/14WVjkqmkhbJQ-IEM_Un0YTTiB2jKJqQD?usp=sharing). Models are available for MipNerf-360, Tanks and Temples and DeepBlending dataset for both CompGS-4k and CompGS-32k variants. 
+
 **[4 January 2024]:** We have updated the code to store and load the quantized parameters using the indices and codebook. Only the compressed version of the model is saved.  
 
 ## Getting Started