pycaffe関連のエラー：Exception: Input blob arguments do not match net inputs.

ここを参考に，caffemodelを自分用に適用してウェブカメラ映像からリアルタイムで人を検出しようとしています．

このスクリプトを実行したところ，このようなエラーが出てきました．

Traceback (most recent call last):
  File "realtime2.py", line 150, in <module>
    [im2, cls, dets, CONF_THRESH] = demo(net, frame, scale_factor, ('person',))
  File "realtime2.py", line 87, in demo
    scores, boxes = im_detect(net, im2, obj_proposals)
  File "/home/keisan/py-faster-rcnn/tools/../lib/fast_rcnn/test.py", line 154, in im_detect
    blobs_out = net.forward(**forward_kwargs)
  File "/home/keisan/py-faster-rcnn/tools/../caffe-fast-rcnn/python/caffe/pycaffe.py", line 97, in _Net_forward
    raise Exception('Input blob arguments do not match net inputs.')
Exception: Input blob arguments do not match net inputs.

このエラーの意味はどういった内容なのでしょうか？（ググってもいまいちピンときません）
また，どのような対処法が考えられるでしょうか？
どなたかご教示よろしくお願いします．

私のスクリプトは以下のとおりです

python
1#!/usr/bin/env python
2# -*- coding: utf-8 -*-
3
4# --------------------------------------------------------
5# Fast R-CNN
6# Copyright (c) 2015 Microsoft
7# Licensed under The MIT License [see LICENSE for details]
8# Written by Ross Girshick
9# --------------------------------------------------------
10
11"""
12Demo script showing detections in sample images.
13See README.md for installation instructions before running.
14"""
15
16import _init_paths
17from fast_rcnn.config import cfg
18from fast_rcnn.test import im_detect
19from fast_rcnn.nms_wrapper import nms
20from utils.timer import Timer
21import matplotlib.pyplot as plt
22import numpy as np
23import scipy.io as sio
24import caffe, os, cv2 
25import argparse
26import dlib
27
28CLASSES = ('__background__',
29           'aeroplane', 'bicycle', 'bird', 'boat',
30           'bottle', 'bus', 'car', 'cat', 'chair',
31           'cow', 'diningtable', 'dog', 'horse',
32           'motorbike', 'person', 'pottedplant',
33           'sheep', 'sofa', 'train', 'tvmonitor')
34
35NETS = {'vgg16': ('VGG16',
36                  'VGG16_faster_rcnn_final.caffemodel'),
37        'zf': ('ZF',
38                    'ZF_faster_rcnn_final.caffemodel')}
39
40
41def vis_detections(im, class_name, dets, thresh=0.5):
42    """Draw detected bounding boxes."""
43    inds = np.where(dets[:, -1] >= thresh)[0]
44    if len(inds) == 0:
45        return
46
47    im = im[:, :, (2, 1, 0)]
48    fig, ax = plt.subplots(figsize=(12, 12))
49    ax.imshow(im, aspect='equal')
50    for i in inds:
51        bbox = dets[i, :4]
52        score = dets[i, -1]
53
54        ax.add_patch(
55            plt.Rectangle((bbox[0], bbox[1]),
56                          bbox[2] - bbox[0],
57                          bbox[3] - bbox[1], fill=False,
58                          edgecolor='red', linewidth=3.5)
59            )
60        ax.text(bbox[0], bbox[1] - 2,
61                '{:s} {:.3f}'.format(class_name, score),
62                bbox=dict(facecolor='blue', alpha=0.5),
63                fontsize=14, color='white')
64
65    ax.set_title(('{} detections with '
66                  'p({} | box) >= {:.1f}').format(class_name, class_name,
67                                                  thresh),
68                  fontsize=14)
69    plt.axis('off')
70    plt.tight_layout()
71    plt.draw()
72
73def demo(net, im, scale_factor, classes):
74    """Detect object classes in an image using pre-computed object proposals."""
75
76    im2 = cv2.resize(im, (0,0), fx=1.0/scale_factor, fy=1.0/scale_factor)
77
78    obj_proposals_in = []
79    dlib.find_candidate_object_locations(im2, obj_proposals_in, min_size=70)
80
81    obj_proposals = np.empty((len(obj_proposals_in),4))
82    for idx in range(len(obj_proposals_in)):
83        obj_proposals[idx] = [obj_proposals_in[idx].left(), obj_proposals_in[idx].top(), obj_proposals_in[idx].right(), obj_proposals_in[idx].bottom()]
84
85    # Detect all object classes and regress object bounds
86    scores, boxes = im_detect(net, im2, obj_proposals)
87
88    # Visualize detections for each class
89    CONF_THRESH = 0.8
90    NMS_THRESH = 0.3
91    for cls in classes:
92        cls_ind = CLASSES.index(cls)
93        cls_boxes = boxes[:, 4*cls_ind:4*(cls_ind + 1)]
94        cls_scores = scores[:, cls_ind]
95        dets = np.hstack((cls_boxes,
96                          cls_scores[:, np.newaxis])).astype(np.float32)
97        keep = nms(dets, NMS_THRESH)
98        dets = dets[keep, :]
99
100    return [im2, cls, dets, CONF_THRESH]
101
102
103def parse_args():
104    """Parse input arguments."""
105    parser = argparse.ArgumentParser(description='Train a Fast R-CNN network')
106    parser.add_argument('--gpu', dest='gpu_id', help='GPU device id to use [0]',
107                        default=0, type=int)
108    parser.add_argument('--cpu', dest='cpu_mode',
109                        help='Use CPU mode (overrides --gpu)',
110                        action='store_true')
111    parser.add_argument('--net', dest='demo_net', help='Network to use [vgg16]',
112                        choices=NETS.keys(), default='vgg16')
113
114    args = parser.parse_args()
115
116    return args
117
118
119if __name__ == '__main__':
120    args = parse_args()
121
122    prototxt = os.path.join(cfg.MODELS_DIR, NETS[args.demo_net][0], 
123                            'faster_rcnn_alt_opt', 'faster_rcnn_test.pt')
124    caffemodel = os.path.join(cfg.DATA_DIR, 'faster_rcnn_models',
125                              NETS[args.demo_net][1])
126
127    if not os.path.isfile(caffemodel):
128        raise IOError(('{:s} not found.\nDid you run ./data/script/'
129                       'fetch_fast_rcnn_models.sh?').format(caffemodel))
130
131    if args.cpu_mode:
132        caffe.set_mode_cpu()
133    else:
134        caffe.set_mode_gpu()
135    caffe.set_device(args.gpu_id)
136    net = caffe.Net(prototxt, caffemodel, caffe.TEST)
137
138    print '\n\nLoaded network {:s}'.format(caffemodel)
139
140    cap = cv2.VideoCapture(0)
141
142    while(True):
143        # Capture frame-by-frame
144        ret, frame = cap.read()
145
146        # Scaling the video feed can help the system run faster (and run on GPUs with less memory)
147        # e.g. with a standard video stream of size 640x480, a scale_factor = 4 will allow the system to run a < 1 sec/frame
148        scale_factor = 4
149        [im2, cls, dets, CONF_THRESH] = demo(net, frame, scale_factor, ('person',))
150
151        inds = np.where(dets[:, -1] >= CONF_THRESH)[0]
152        if len(inds) != 0:
153            for i in inds:
154                bbox = dets[i, :4]
155                cv2.rectangle(frame,(int(bbox[0]*scale_factor),int(bbox[1]*scale_factor)),(int(bbox[2]*scale_factor),int(bbox[3]*scale_factor)),(0,255,0),2)
156
157        # Display the resulting frame
158        cv2.imshow('frame',frame)
159
160        if cv2.waitKey(1) & 0xFF == ord('q'):
161            break
162
163    # When everything done, release the capture
164    cap.release()
165cv2.destroyAllWindows()