diff --git a/datasets/make_datasets.py b/datasets/make_datasets.py
index c5b63b0..1194d9a 100644
--- a/datasets/make_datasets.py
+++ b/datasets/make_datasets.py
@@ -137,7 +137,8 @@
 
     with open(osp.join(options.out_dir, "environment.yaml"), "w") as f:
         dataset_info = {"height": options.save_height,
-                        "width": options.save_width}
+                        "width": options.save_width,
+                        "frame_freq": options.save_frequency}
         camera_info = {"intrinsic": intrinsic}
         environment = {"dataset_info": dataset_info,
                        "camera_info": camera_info}
diff --git a/prediction/pred_disp.py b/prediction/pred_disp.py
index 443efd2..f091123 100644
--- a/prediction/pred_disp.py
+++ b/prediction/pred_disp.py
@@ -23,7 +23,7 @@
 
 device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
 disp_net = models.DispResNet(18, False).to(device)
-disp_net = torch.nn.DataParallel(disp_net)
+# disp_net = torch.nn.DataParallel(disp_net)
 weights = torch.load(args.weight_pth)
 disp_net.load_state_dict(weights)
 disp_net.eval()
diff --git a/prediction/pred_pose.py b/prediction/pred_pose.py
index e170828..f6aa77a 100644
--- a/prediction/pred_pose.py
+++ b/prediction/pred_pose.py
@@ -8,11 +8,88 @@
 import pandas as pd
 import argparse
 import models
+import yaml
+from tqdm import tqdm
+from glob import glob
 
 parser = argparse.ArgumentParser(description='Structure from Motion Learner',
                                  formatter_class=argparse.ArgumentDefaultsHelpFormatter)
 
 parser.add_argument('--weight_pth', required=True, help="path to disp model's pth file")
 parser.add_argument('--path_to_yaml', default="../datasets/data_for_SC_SfMLearner/environment.yaml", help="path to environent")
-parser.add_argument('--use_camera', type=int, default=-9999, help="カメラを使うならポート番号を指定")
+parser.add_argument('--use_camera', type=int, default=-9999, help="カメラをq使うならポート番号を指定")
 args = parser.parse_args()
+
+device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
+
+with open(args.path_to_yaml) as f:
+    environment = yaml.safe_load(f)
+r_h, r_w = environment["dataset_info"]["height"], environment["dataset_info"]["width"]
+pass_num = environment["dataset_info"]["frame_freq"]
+
+def cvmat2tensor(cvmat):
+    cvmat = cv2.cvtColor(cvmat, cv2.COLOR_BGR2RGB)
+    cvmat = np.transpose(cvmat, (2, 0, 1)).astype(np.float32)
+    input_tensor = torch.from_numpy(cvmat).unsqueeze(0)
+    input_tensor = ((input_tensor / 255 - 0.45) / 0.225).to(device)
+
+    return input_tensor
+
+def run_inference_pose(cap, save_name, pose_net):
+    print("start processing => {}".format(save_name))
+    ret, prev_frame = cap.read()
+    prev_frame = cv2.resize(prev_frame, (r_w, r_h), interpolation=cv2.INTER_LINEAR)
+    prev_tensor = cvmat2tensor(prev_frame)
+
+    iter_num = 0
+    save_poses = list()
+    while True:
+        ret, cur_frame = cap.read()
+
+        if not ret:
+            break
+
+        if cv2.waitKey(1) & 0xFF == ord('q'):
+            break
+
+        cur_frame = cv2.resize(cur_frame, (r_w, r_h), interpolation=cv2.INTER_LINEAR)
+        if iter_num % pass_num == (pass_num - 1):
+            cur_tensor = cvmat2tensor(cur_frame)
+            poses = pose_net(cur_tensor, prev_tensor)[0]
+            poses = [item.cpu().item() for item in list(poses)]
+            print(poses)
+            save_poses.append(poses)
+            prev_tensor = cur_tensor
+
+        cv2.imshow("current frame", cur_frame)
+        cv2.waitKey(1)
+        iter_num += 1
+
+    print("finish processing => {}".format(save_name))
+    save_df = pd.DataFrame(save_poses)
+    save_df.columns = ["tx", "ty", "tz", "Rx", "Ry", "Rz"]
+    save_df.to_csv("./output/{}.csv".format(save_name), index=False)
+
+
+
+pose_net = models.PoseResNet(18, False).to(device)
+pose_net = torch.nn.DataParallel(pose_net)
+weights = torch.load(args.weight_pth)
+pose_net.load_state_dict(weights)
+pose_net.eval()
+print("カメラのポーズは[tx, ty, tz, Rx, Ry, Rz]のように表示されます．")
+
+if args.use_camera != -9999:
+    print("「q」キーを押すことでプログラムが終了します．")
+    cap = cv2.VideoCapture(args.use_camera)
+    run_inference_pose(cap, "usb", pose_net)
+else:
+    target_file_list = glob(os.path.join("input", "*.*"))
+
+    for target_file in tqdm(target_file_list):
+        ext = target_file.lower().split('.')[-1]
+
+        if ext == "mp4":
+            cap = cv2.VideoCapture(target_file)
+            save_name = os.path.splitext(os.path.basename(target_file))[0]
+            run_inference_pose(cap, save_name, pose_net)
diff --git a/requirements.txt b/requirements.txt
index d81625f..67684cc 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,5 +1,6 @@
 PyYAML==5.3.1
 tensorboardX==2.1
+tensorboard==2.3.0
 opencv-python-headless==4.4.0.40
 path==15.0.0
 matplotlib==3.3.1
diff --git a/train.py b/train.py
index 8b4c38c..5cf6bb4 100644
--- a/train.py
+++ b/train.py
@@ -85,6 +85,7 @@
         environment = yaml.safe_load(f)
 
     timestamp = Path(datetime.datetime.now().strftime("%m-%d-%H-%M"))
+    os.makedirs("./checkpoints", exist_ok=True)
     args.save_path = 'checkpoints' / timestamp
     print('=> will save everything to {}'.format(args.save_path))
     args.save_path.makedirs_p()