diff --git a/demo/__init__.py b/demo/__init__.py
new file mode 100644
index 0000000..9ab16e0
--- /dev/null
+++ b/demo/__init__.py
@@ -0,0 +1,8 @@
+# -*- coding: utf-8 -*-
+
+"""
+@date: 2020/10/30 下午3:22
+@file: __init__.py.py
+@author: zj
+@description: 
+"""
diff --git a/demo/tsn_r50_ucf101_rgb_raw_dense_1x16x4.yaml b/demo/tsn_r50_ucf101_rgb_raw_dense_1x16x4.yaml
index 36c87fb..400c438 100644
--- a/demo/tsn_r50_ucf101_rgb_raw_dense_1x16x4.yaml
+++ b/demo/tsn_r50_ucf101_rgb_raw_dense_1x16x4.yaml
@@ -3,12 +3,14 @@ NUM_NODES: 1
 RANK_ID: 0
 DIST_BACKEND: "nccl"
 RNG_SEED: 1
-OUTPUT_DIR: 'outputs/tsn_r50_ucf101_rgb_raw_dense_1x16x4'
 DATASETS:
   MODALITY: 'RGB'
+  TYPE: 'RawFrame'
+  SAMPLE_STRATEGY: 'DenseSample'
   CLIP_LEN: 1
   FRAME_INTERVAL: 16
   NUM_CLIPS: 4
+  NUM_SAMPLE_POSITIONS: 1
 TRANSFORM:
   MEAN: (0.485, 0.456, 0.406)
   STD: (0.229, 0.224, 0.225)
@@ -21,9 +23,8 @@ MODEL:
   NAME: 'TSN'
   PRETRAINED: ''
   SYNC_BN: True
-  INPUT_SIZE: (224, 224, 3)
   BACKBONE:
-    NAME: 'resnet50'
+    NAME: 'ResNet50'
     PARTIAL_BN: False
     TORCHVISION_PRETRAINED: True
     ZERO_INIT_RESIDUAL: True
@@ -48,4 +49,4 @@ VISUALIZATION:
   DISPLAY_HEIGHT: 0
   OUTPUT_FPS: -1
   OUTPUT_FILE: ""
-  LABEL_FILE_PATH: 'data/ucf101/annotations/classInd.txt'
+  LABEL_FILE_PATH: 'data/ucf101/annotations/classInd.txt'
\ No newline at end of file
diff --git a/tsn/visualization/predictor/predictor.py b/tsn/visualization/predictor/predictor.py
index abf2b98..62c0368 100644
--- a/tsn/visualization/predictor/predictor.py
+++ b/tsn/visualization/predictor/predictor.py
@@ -3,7 +3,7 @@
 
 import torch
 
-from tsn.model.build import build_model
+from tsn.model.recognizers.build import build_recognizer
 from tsn.data.transforms.build import build_transform
 from .util import process_cv2_inputs
 from tsn.util.distributed import get_device, get_local_rank
@@ -27,7 +27,7 @@ def __init__(self, cfg):
             device = get_device()
 
         # Build the video model and print model statistics.
-        self.model = build_model(cfg, device)
+        self.model = build_recognizer(cfg, device)
         self.model.eval()
         self.transform = build_transform(cfg, is_train=False)