Update README.md

README.md

@@ -222,67 +222,67 @@ The following steps give the steps required to train a faster rcnn on
 a hypothetical "anime" dataset.
 
 1. Create a softlink for your dataset folder 
-```
-cd $FRCN_ROOT/data
-ln -s /home/ubuntu/train_data/anime anime
-```
+	```
+	cd $FRCN_ROOT/data
+	ln -s /home/ubuntu/train_data/anime anime
+	```
 
 2. Define a class for reading the dataset
-```
-# e.g
-cp $FRCN_ROOT/lib/datasets/pascal_voc.py $FRCN_ROOT/lib/datasets/anime.py
-
-# update the file and change the number of classes to match your dataset
-```
+	```
+	# e.g
+	cp $FRCN_ROOT/lib/datasets/pascal_voc.py $FRCN_ROOT/lib/datasets/anime.py
+	
+	# update the file and change the number of classes to match your dataset
+	```
 
 3. Define your dataset splits (You will need to import the class defined in step 2.)
-```
-vi $FRCN_ROOT/lib/datasets/factory.py
-
-from datasets.anime import anime
-
-# e.g. anime dataset
- for year in ['2016']:
-     for split in ['train', 'val', 'trainval', 'test']:
- 	name = 'anime_{}_{}'.format(year, split)
- 	
- 	# initialize multiple "anime" objects for different data splits
- 	__sets[name] = (lambda split=split, year=year: anime(split, year))
-```
+	```
+	vi $FRCN_ROOT/lib/datasets/factory.py
+	
+	from datasets.anime import anime
+	
+	# e.g. anime dataset
+	 for year in ['2016']:
+	     for split in ['train', 'val', 'trainval', 'test']:
+	 	name = 'anime_{}_{}'.format(year, split)
+	 	
+	 	# initialize multiple "anime" objects for different data splits
+	 	__sets[name] = (lambda split=split, year=year: anime(split, year))
+	```
 
 4. Create a models directory which stores caffe configuration files
-```
-cp $FRCN_ROOT/models/pascal_voc $FRCN_ROOT/models/anime
-```
+	```
+	cp $FRCN_ROOT/models/pascal_voc $FRCN_ROOT/models/anime
+	```
 
 5. Update the model configs for the pretrained model you want to start your training with.
 E.g. we use the ZF network
-```
-vi $FRCN_ROOT/models/anime/ZF/faster_rcnn_end2end/train.prototxt
-
-# Change the 'num_classes' in layer 'input-data' to match your dataset
-# Change the 'num_output' in layer 'bbox_pred' to match your dataset. Note that bbox_output = 4 * num_classes
-# Change the 'num_output' in layer 'cls_score' to match your dataset.
-```
+	```
+	vi $FRCN_ROOT/models/anime/ZF/faster_rcnn_end2end/train.prototxt
+	
+	# Change the 'num_classes' in layer 'input-data' to match your dataset
+	# Change the 'num_output' in layer 'bbox_pred' to match your dataset. Note that bbox_output = 4 * num_classes
+	# Change the 'num_output' in layer 'cls_score' to match your dataset.
+	```
 
 6. Change the execution script to add your newly created dataset
-```
-vi $FRCN_ROOT/experiments/scripts/faster_rcnn_end2end.sh
-
-# Add your dataset under case $DATASET
-
-anime)
-    TRAIN_IMDB="anime_2007_trainval"
-    TEST_IMDB="anime_2007_test"
-    PT_DIR="anime"
-    ITERS=70000
-  ;;
-```
+	```
+	vi $FRCN_ROOT/experiments/scripts/faster_rcnn_end2end.sh
+	
+	# Add your dataset under case $DATASET
+	
+	anime)
+	    TRAIN_IMDB="anime_2007_trainval"
+	    TEST_IMDB="anime_2007_test"
+	    PT_DIR="anime"
+	    ITERS=70000
+	  ;;
+	```
 
 7. Train the model!
-```
-cd $FRCN_ROOT
-./experiments/scripts/faster_rcnn_end2end.sh 0 ZF anime
-```
+	```
+	cd $FRCN_ROOT
+	./experiments/scripts/faster_rcnn_end2end.sh 0 ZF anime
+	```
 
 If everything goes smoothly, the model should be training.