Updated Readme and requirements.txt

README.md

@@ -1,26 +1,32 @@
 # Shadow Detector
 
-## Installing pymeanshift
+## Installing Dependencies 
 
 ```
+pip install -r requirements.txt
+git submodule update --init
 cd pymeanshift
 ./setup.py install
 ```
-## Generating dataset
+
+## Downloading SBU-Dataset and setting up project structure
 
 ```
-./make_dataset.py
+./init.sh
 ```
 
+Running this will download the [SBU shadow dataset](http://www3.cs.stonybrook.edu/~cvl/content/datasets/shadow_db/SBU-shadow.zip)
+and extract it under the ```data``` directory, as well as create the
+```checkpoints``` directory.
+
 ## Training the model
 
 ```
 ./train.py
 ```
 
-This will save the model along with the weights at ```model.h5``` after the
-final iteration of training and also saves the models with the best accuracy
-under the ```checkpoints``` directory.
+This will save the models with the best accuracies under the ```checkpoints```
+directory.
 
 ## Detecting shadows
 

init.sh

@@ -0,0 +1,12 @@
+mkdir checkpoints
+
+# Download SBU-shadow dataset and rename folder.
+wget http://www3.cs.stonybrook.edu/~cvl/content/datasets/shadow_db/SBU-shadow.zip
+unzip SBU-shadow.zip -d ./
+mv SBU-shadow data
+rm SBU-shadow.zip
+
+rm ./data/SBUTrain4KRecoveredSmall/ShadowImages/*.mat
+rm ./data/SBUTrain4KRecoveredSmall/ShadowMasks/*.mat
+rm ./data/SBU-Test/ShadowImages/*.mat
+rm ./data/SBU-Test/ShadowMasks/*.mat

requirements.txt

@@ -12,7 +12,6 @@ Markdown==2.6.11
 numpy==1.14.5
 opencv-python==3.4.1.15
 protobuf==3.6.0
-pymeanshift==0.2.1
 PyYAML==3.12
 scipy==1.1.0
 six==1.11.0