diff --git a/code/ensemble.py b/code/ensemble.py
new file mode 100644
index 0000000..cafbf27
--- /dev/null
+++ b/code/ensemble.py
@@ -0,0 +1,43 @@
+import pandas as pd
+import numpy as np
+import glob
+from tqdm import tqdm
+
+data_root = 'ensemble'
+
+df_list = []
+for path in glob.glob(data_root+'/*.csv'):
+    df = pd.read_csv(path)
+    df_list.append(df)
+
+scores_list = []
+for path in glob.glob(data_root+'/*.txt'):
+    with open(path,'r') as f:
+        scores = f.readline()
+        scores = list(map(float,scores.split(',')))
+    scores_list.append(scores)
+
+def get_encoding(img,score):
+    img_list = img.split(' ')
+    img_np = np.array(img_list,dtype=int)
+    encoding = np.eye(11)[img_np]
+    encoding *= score
+    return encoding
+
+
+out_dict = {'image_id':[],'PredictionString':[]}
+for i in tqdm(range(len(df_list[0]))):
+    out_dict['image_id'].append(df_list[0]['image_id'][i])
+    
+    encoding = np.zeros((65536,11))
+    for df,scores in zip(df_list,scores_list):
+        encoding += get_encoding(df['PredictionString'][i],scores)
+
+    out_img_np = np.argmax(encoding,axis=1)
+    out_img_list = out_img_np.tolist()
+    out_img = ' '.join(map(str,out_img_list))
+    out_dict['PredictionString'].append(out_img)
+
+
+out_df = pd.DataFrame(out_dict)
+out_df.to_csv('ensemble.csv',index=False)
\ No newline at end of file