diff --git a/CMakeLists.txt b/CMakeLists.txt
index d63891d..15702bd 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -9,7 +9,7 @@ endif()
 set(HAVE_LIBPTHREAD YES)
 
 add_definitions(-std=c++11)
-set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11 -pthread")
 
 include_directories(/usr/local/include)
 include_directories(..)
diff --git a/README.md b/README.md
index f7a88ff..1dda6ba 100644
--- a/README.md
+++ b/README.md
@@ -33,9 +33,7 @@ Recognize input sketch image, you can try it via [web version](http://online.ope
 How to install
 ==============
 
-OpenSSE core library only depends on [OpenCV 3.0.0+](http://opencv.org/), it's a cross-platform library.
-
-You can compile and install it like below:
+After installed OpenSSE's dependency - [OpenCV 3.0.0+](http://opencv.org/), You can compile and install it like below:
 
 ```sh
 mkdir release && cd release
@@ -52,8 +50,6 @@ More detail info, go to [OpenSSE Wiki](https://github.com/zddhub/opensse/wiki)
 
 Todo List
 =========
-- [x] Docker support
-- [x] Remove boost for core library, done
 - [ ] Speed up train steps
 - [ ] Recognize sketch
 - [ ] Support big image set
diff --git a/sse/vocabulary/kmeans.h b/sse/vocabulary/kmeans.h
index 413ec05..9ce9453 100644
--- a/sse/vocabulary/kmeans.h
+++ b/sse/vocabulary/kmeans.h
@@ -89,9 +89,7 @@ class Kmeans
             std::size_t changes = 0;
 
             // distribute items on clusters in parallel
-            // thread_group pool;
             std::vector<std::thread> pools(std::thread::hardware_concurrency());
-            // pools.emplace_back(functor); // pass in the argument of std::thread()
 
             std::size_t idx = 0;
             mutex_t mtx;