Highlight NOTE

README.md

@@ -41,7 +41,7 @@ conda install pytorch torchvision pytorch-cuda=12.1 -c pytorch -c nvidia
 pip install -e .
 ```
 
-> **NOTE:**
+> [!NOTE]
 > - We use an adjusted version of [nvdiffrast](https://github.com/ShenhanQian/nvdiffrast/tree/backface-culling) for backface-culling. To completely remove previous versions and compiled pytorch extensions, you can execute
 >
 >   ```shell
@@ -60,7 +60,8 @@ Our code relies on FLAME. Please download assets from the [official website](htt
 - FLAME 2023 (versions w/ jaw rotation) -> `asset/flame/flame2023.pkl`
 - FLAME Vertex Masks -> `asset/flame/FLAME_masks.pkl`
 
-> **NOTE:** It is possible to use FLAME 2020 by download to `asset/flame/generic_model.pkl`. The `FLAME_MODEL_PATH` in `flame.py` needs to be updated accordingly.
+> [!NOTE]
+> It is possible to use FLAME 2020 by download to `asset/flame/generic_model.pkl`. The `FLAME_MODEL_PATH` in `flame.py` needs to be updated accordingly.
 
 ### Video Data
 

doc/nersemble.md

@@ -46,7 +46,8 @@ Optional arguments
 
 - `--exp.no_photometric`: track only with landmark (very fast, but coarse)
 
-> **NOTE:** We use all 16 views for the optimization, but we only visualize 3 views for efficiency.
+> [!NOTE]
+> We use all 16 views for the optimization, but we only visualize 3 views for efficiency.
 
 ### 3. Export tracking results into a NeRF-style dataset
 
@@ -84,4 +85,5 @@ python vhap/combine_nerf_datasets.py \
   export/nersemble/UNION10_${SUBJECT}_EMO1234EXP234589_v16_DS4_whiteBg_staticOffset_maskBelowLine
 ```
 
-> **NOTE**: The `tgt_folder` must be in the same parent folder as `src_folders` because the union dataset read from the original image files by relative paths.
+> [!NOTE]
+> The `tgt_folder` must be in the same parent folder as `src_folders` because the union dataset read from the original image files by relative paths.