diff --git a/README.md b/README.md
index fe7530176..f561ede98 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# Faster R-CNN for KITTI and BDD100k
+# Faster R-CNN for KITTI and BDD100K
 
 This project is based on facebookresearch's [maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark).
 
@@ -30,7 +30,7 @@ pip install tensorboardX
 ### KITTI
 This repo assumes that you followed NVIDIA DIGIT's instruction to prepare the KITTI dataset.
 
-### BDD100k
+### BDD100K
 Prepare the dataset with the below structure:
 ```
 bdd100k/
@@ -56,12 +56,12 @@ ln -s {parent directory of kitti and bdd100k} {root directory of this repo}/data
 python tools/train_net.py --config-file "configs/e2e_faster_rcnn_R_101_FPN_1x.yaml" --use_tensorboard MODEL.ROI_BOX_HEAD.NUM_CLASSES 7 DATASETS.TRAIN '("kitti_train",)' OUTPUT_DIR out_kitti
 ```
 
-### BDD100k
+### BDD100K
 ```
 python tools/train_net.py --config-file "configs/e2e_faster_rcnn_R_101_FPN_1x.yaml" --use_tensorboard MODEL.ROI_BOX_HEAD.NUM_CLASSES 11 DATASETS.TRAIN '("bdd100k_train",)' OUTPUT_DIR out_bdd
 ```
 
-For the BDD100k dataset, you might need to clip the gradient for training stability. For other hyperparameters, refer to [maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark).
+For the BDD100K dataset, you might need to clip the gradient for training stability. For other hyperparameters, refer to [maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark).
 
 
 ## Inference
@@ -73,7 +73,7 @@ Below commands loads saved weights and save the detections in the [BDD format](h
 python tools/test_net.py --weights out_kitti/*.pth --config-file "configs/e2e_faster_rcnn_R_101_FPN_1x.yaml" DATASETS.TEST '("kitti_val",)' OUTPUT_DIR out_kitti
 ```
 
-### BDD100k
+### BDD100K
 ```
 python tools/test_net.py --weights out_bdd/*.pth --config-file "configs/e2e_faster_rcnn_R_101_FPN_1x.yaml" DATASETS.TEST '("bdd100k_val",)' OUTPUT_DIR out_bdd
 ```