Hugging Face's logo

Datasets:
hexuan21
/
VideoScore_legacy

License:
Dataset card Data Studio Files
xet
Community
hexuan21 commited on
Commit
75d4be8
·
verified ·
1 Parent(s): 2f884e9

upload refined_40k/test_bench 's json, videos and frames

Browse files
Files changed (1) hide show
  1. data/refined_40k/test_bench/data_fastsvd.json +782 -0
data/refined_40k/test_bench/data_fastsvd.json ADDED
@@ -0,0 +1,782 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "id": "9000913",
4
+ "images": [
5
+ "9000913_00.jpg",
6
+ "9000913_01.jpg",
7
+ "9000913_02.jpg",
8
+ "9000913_03.jpg",
9
+ "9000913_04.jpg",
10
+ "9000913_05.jpg",
11
+ "9000913_06.jpg",
12
+ "9000913_07.jpg",
13
+ "9000913_08.jpg",
14
+ "9000913_09.jpg",
15
+ "9000913_10.jpg",
16
+ "9000913_11.jpg",
17
+ "9000913_12.jpg",
18
+ "9000913_13.jpg",
19
+ "9000913_14.jpg",
20
+ "9000913_15.jpg",
21
+ "9000913_16.jpg",
22
+ "9000913_17.jpg",
23
+ "9000913_18.jpg",
24
+ "9000913_19.jpg",
25
+ "9000913_20.jpg",
26
+ "9000913_21.jpg",
27
+ "9000913_22.jpg",
28
+ "9000913_23.jpg"
29
+ ],
30
+ "conversations": [
31
+ {
32
+ "from": "human",
33
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"windmill\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
34
+ },
35
+ {
36
+ "from": "gpt",
37
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
38
+ }
39
+ ]
40
+ },
41
+ {
42
+ "id": "9000754",
43
+ "images": [
44
+ "9000754_00.jpg",
45
+ "9000754_01.jpg",
46
+ "9000754_02.jpg",
47
+ "9000754_03.jpg",
48
+ "9000754_04.jpg",
49
+ "9000754_05.jpg",
50
+ "9000754_06.jpg",
51
+ "9000754_07.jpg",
52
+ "9000754_08.jpg",
53
+ "9000754_09.jpg",
54
+ "9000754_10.jpg",
55
+ "9000754_11.jpg",
56
+ "9000754_12.jpg",
57
+ "9000754_13.jpg",
58
+ "9000754_14.jpg",
59
+ "9000754_15.jpg",
60
+ "9000754_16.jpg",
61
+ "9000754_17.jpg",
62
+ "9000754_18.jpg",
63
+ "9000754_19.jpg",
64
+ "9000754_20.jpg",
65
+ "9000754_21.jpg",
66
+ "9000754_22.jpg",
67
+ "9000754_23.jpg"
68
+ ],
69
+ "conversations": [
70
+ {
71
+ "from": "human",
72
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Yellow flowers swing in the wind\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
73
+ },
74
+ {
75
+ "from": "gpt",
76
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
77
+ }
78
+ ]
79
+ },
80
+ {
81
+ "id": "9000025",
82
+ "images": [
83
+ "9000025_00.jpg",
84
+ "9000025_01.jpg",
85
+ "9000025_02.jpg",
86
+ "9000025_03.jpg",
87
+ "9000025_04.jpg",
88
+ "9000025_05.jpg",
89
+ "9000025_06.jpg",
90
+ "9000025_07.jpg",
91
+ "9000025_08.jpg",
92
+ "9000025_09.jpg",
93
+ "9000025_10.jpg",
94
+ "9000025_11.jpg",
95
+ "9000025_12.jpg",
96
+ "9000025_13.jpg",
97
+ "9000025_14.jpg",
98
+ "9000025_15.jpg",
99
+ "9000025_16.jpg",
100
+ "9000025_17.jpg",
101
+ "9000025_18.jpg",
102
+ "9000025_19.jpg",
103
+ "9000025_20.jpg",
104
+ "9000025_21.jpg",
105
+ "9000025_22.jpg",
106
+ "9000025_23.jpg"
107
+ ],
108
+ "conversations": [
109
+ {
110
+ "from": "human",
111
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a bowl and a remote\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
112
+ },
113
+ {
114
+ "from": "gpt",
115
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
116
+ }
117
+ ]
118
+ },
119
+ {
120
+ "id": "9000142",
121
+ "images": [
122
+ "9000142_00.jpg",
123
+ "9000142_01.jpg",
124
+ "9000142_02.jpg",
125
+ "9000142_03.jpg",
126
+ "9000142_04.jpg",
127
+ "9000142_05.jpg",
128
+ "9000142_06.jpg",
129
+ "9000142_07.jpg",
130
+ "9000142_08.jpg",
131
+ "9000142_09.jpg",
132
+ "9000142_10.jpg",
133
+ "9000142_11.jpg",
134
+ "9000142_12.jpg",
135
+ "9000142_13.jpg",
136
+ "9000142_14.jpg",
137
+ "9000142_15.jpg",
138
+ "9000142_16.jpg",
139
+ "9000142_17.jpg",
140
+ "9000142_18.jpg",
141
+ "9000142_19.jpg",
142
+ "9000142_20.jpg",
143
+ "9000142_21.jpg",
144
+ "9000142_22.jpg",
145
+ "9000142_23.jpg"
146
+ ],
147
+ "conversations": [
148
+ {
149
+ "from": "human",
150
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"An astronaut is riding a horse in the space in a photorealistic style\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
151
+ },
152
+ {
153
+ "from": "gpt",
154
+ "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 2\n text-to-video alignment: 3\n factual consistency: 2\n"
155
+ }
156
+ ]
157
+ },
158
+ {
159
+ "id": "9000104",
160
+ "images": [
161
+ "9000104_00.jpg",
162
+ "9000104_01.jpg",
163
+ "9000104_02.jpg",
164
+ "9000104_03.jpg",
165
+ "9000104_04.jpg",
166
+ "9000104_05.jpg",
167
+ "9000104_06.jpg",
168
+ "9000104_07.jpg",
169
+ "9000104_08.jpg",
170
+ "9000104_09.jpg",
171
+ "9000104_10.jpg",
172
+ "9000104_11.jpg",
173
+ "9000104_12.jpg",
174
+ "9000104_13.jpg",
175
+ "9000104_14.jpg",
176
+ "9000104_15.jpg",
177
+ "9000104_16.jpg",
178
+ "9000104_17.jpg",
179
+ "9000104_18.jpg",
180
+ "9000104_19.jpg",
181
+ "9000104_20.jpg",
182
+ "9000104_21.jpg",
183
+ "9000104_22.jpg",
184
+ "9000104_23.jpg"
185
+ ],
186
+ "conversations": [
187
+ {
188
+ "from": "human",
189
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"An astronaut flying in space, in cyberpunk style\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
190
+ },
191
+ {
192
+ "from": "gpt",
193
+ "value": "visual quality: 3\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
194
+ }
195
+ ]
196
+ },
197
+ {
198
+ "id": "9000089",
199
+ "images": [
200
+ "9000089_00.jpg",
201
+ "9000089_01.jpg",
202
+ "9000089_02.jpg",
203
+ "9000089_03.jpg",
204
+ "9000089_04.jpg",
205
+ "9000089_05.jpg",
206
+ "9000089_06.jpg",
207
+ "9000089_07.jpg",
208
+ "9000089_08.jpg",
209
+ "9000089_09.jpg",
210
+ "9000089_10.jpg",
211
+ "9000089_11.jpg",
212
+ "9000089_12.jpg",
213
+ "9000089_13.jpg",
214
+ "9000089_14.jpg",
215
+ "9000089_15.jpg",
216
+ "9000089_16.jpg",
217
+ "9000089_17.jpg",
218
+ "9000089_18.jpg",
219
+ "9000089_19.jpg",
220
+ "9000089_20.jpg",
221
+ "9000089_21.jpg",
222
+ "9000089_22.jpg",
223
+ "9000089_23.jpg"
224
+ ],
225
+ "conversations": [
226
+ {
227
+ "from": "human",
228
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a pink suitcase\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
229
+ },
230
+ {
231
+ "from": "gpt",
232
+ "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
233
+ }
234
+ ]
235
+ },
236
+ {
237
+ "id": "9000758",
238
+ "images": [
239
+ "9000758_00.jpg",
240
+ "9000758_01.jpg",
241
+ "9000758_02.jpg",
242
+ "9000758_03.jpg",
243
+ "9000758_04.jpg",
244
+ "9000758_05.jpg",
245
+ "9000758_06.jpg",
246
+ "9000758_07.jpg",
247
+ "9000758_08.jpg",
248
+ "9000758_09.jpg",
249
+ "9000758_10.jpg",
250
+ "9000758_11.jpg",
251
+ "9000758_12.jpg",
252
+ "9000758_13.jpg",
253
+ "9000758_14.jpg",
254
+ "9000758_15.jpg",
255
+ "9000758_16.jpg",
256
+ "9000758_17.jpg",
257
+ "9000758_18.jpg",
258
+ "9000758_19.jpg",
259
+ "9000758_20.jpg",
260
+ "9000758_21.jpg",
261
+ "9000758_22.jpg",
262
+ "9000758_23.jpg"
263
+ ],
264
+ "conversations": [
265
+ {
266
+ "from": "human",
267
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A boat sailing leisurely along the Seine River with the Eiffel Tower in background by Vincent van Gogh\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
268
+ },
269
+ {
270
+ "from": "gpt",
271
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n"
272
+ }
273
+ ]
274
+ },
275
+ {
276
+ "id": "9000030",
277
+ "images": [
278
+ "9000030_00.jpg",
279
+ "9000030_01.jpg",
280
+ "9000030_02.jpg",
281
+ "9000030_03.jpg",
282
+ "9000030_04.jpg",
283
+ "9000030_05.jpg",
284
+ "9000030_06.jpg",
285
+ "9000030_07.jpg",
286
+ "9000030_08.jpg",
287
+ "9000030_09.jpg",
288
+ "9000030_10.jpg",
289
+ "9000030_11.jpg",
290
+ "9000030_12.jpg",
291
+ "9000030_13.jpg",
292
+ "9000030_14.jpg",
293
+ "9000030_15.jpg",
294
+ "9000030_16.jpg",
295
+ "9000030_17.jpg",
296
+ "9000030_18.jpg",
297
+ "9000030_19.jpg",
298
+ "9000030_20.jpg",
299
+ "9000030_21.jpg",
300
+ "9000030_22.jpg",
301
+ "9000030_23.jpg"
302
+ ],
303
+ "conversations": [
304
+ {
305
+ "from": "human",
306
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A person is playing harp\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
307
+ },
308
+ {
309
+ "from": "gpt",
310
+ "value": "visual quality: 2\n temporal consistency: 2\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
311
+ }
312
+ ]
313
+ },
314
+ {
315
+ "id": "9000604",
316
+ "images": [
317
+ "9000604_00.jpg",
318
+ "9000604_01.jpg",
319
+ "9000604_02.jpg",
320
+ "9000604_03.jpg",
321
+ "9000604_04.jpg",
322
+ "9000604_05.jpg",
323
+ "9000604_06.jpg",
324
+ "9000604_07.jpg",
325
+ "9000604_08.jpg",
326
+ "9000604_09.jpg",
327
+ "9000604_10.jpg",
328
+ "9000604_11.jpg",
329
+ "9000604_12.jpg",
330
+ "9000604_13.jpg",
331
+ "9000604_14.jpg",
332
+ "9000604_15.jpg",
333
+ "9000604_16.jpg",
334
+ "9000604_17.jpg",
335
+ "9000604_18.jpg",
336
+ "9000604_19.jpg",
337
+ "9000604_20.jpg",
338
+ "9000604_21.jpg",
339
+ "9000604_22.jpg",
340
+ "9000604_23.jpg"
341
+ ],
342
+ "conversations": [
343
+ {
344
+ "from": "human",
345
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The bund Shanghai, black and white\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
346
+ },
347
+ {
348
+ "from": "gpt",
349
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n"
350
+ }
351
+ ]
352
+ },
353
+ {
354
+ "id": "9000281",
355
+ "images": [
356
+ "9000281_00.jpg",
357
+ "9000281_01.jpg",
358
+ "9000281_02.jpg",
359
+ "9000281_03.jpg",
360
+ "9000281_04.jpg",
361
+ "9000281_05.jpg",
362
+ "9000281_06.jpg",
363
+ "9000281_07.jpg",
364
+ "9000281_08.jpg",
365
+ "9000281_09.jpg",
366
+ "9000281_10.jpg",
367
+ "9000281_11.jpg",
368
+ "9000281_12.jpg",
369
+ "9000281_13.jpg",
370
+ "9000281_14.jpg",
371
+ "9000281_15.jpg",
372
+ "9000281_16.jpg",
373
+ "9000281_17.jpg",
374
+ "9000281_18.jpg",
375
+ "9000281_19.jpg",
376
+ "9000281_20.jpg",
377
+ "9000281_21.jpg",
378
+ "9000281_22.jpg",
379
+ "9000281_23.jpg"
380
+ ],
381
+ "conversations": [
382
+ {
383
+ "from": "human",
384
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a horse and a sheep\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
385
+ },
386
+ {
387
+ "from": "gpt",
388
+ "value": "visual quality: 1\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
389
+ }
390
+ ]
391
+ },
392
+ {
393
+ "id": "9000654",
394
+ "images": [
395
+ "9000654_00.jpg",
396
+ "9000654_01.jpg",
397
+ "9000654_02.jpg",
398
+ "9000654_03.jpg",
399
+ "9000654_04.jpg",
400
+ "9000654_05.jpg",
401
+ "9000654_06.jpg",
402
+ "9000654_07.jpg",
403
+ "9000654_08.jpg",
404
+ "9000654_09.jpg",
405
+ "9000654_10.jpg",
406
+ "9000654_11.jpg",
407
+ "9000654_12.jpg",
408
+ "9000654_13.jpg",
409
+ "9000654_14.jpg",
410
+ "9000654_15.jpg",
411
+ "9000654_16.jpg",
412
+ "9000654_17.jpg",
413
+ "9000654_18.jpg",
414
+ "9000654_19.jpg",
415
+ "9000654_20.jpg",
416
+ "9000654_21.jpg",
417
+ "9000654_22.jpg",
418
+ "9000654_23.jpg"
419
+ ],
420
+ "conversations": [
421
+ {
422
+ "from": "human",
423
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A beautiful coastal beach in spring, waves lapping on sand, with an intense shaking effect\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
424
+ },
425
+ {
426
+ "from": "gpt",
427
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
428
+ }
429
+ ]
430
+ },
431
+ {
432
+ "id": "9000228",
433
+ "images": [
434
+ "9000228_00.jpg",
435
+ "9000228_01.jpg",
436
+ "9000228_02.jpg",
437
+ "9000228_03.jpg",
438
+ "9000228_04.jpg",
439
+ "9000228_05.jpg",
440
+ "9000228_06.jpg",
441
+ "9000228_07.jpg",
442
+ "9000228_08.jpg",
443
+ "9000228_09.jpg",
444
+ "9000228_10.jpg",
445
+ "9000228_11.jpg",
446
+ "9000228_12.jpg",
447
+ "9000228_13.jpg",
448
+ "9000228_14.jpg",
449
+ "9000228_15.jpg",
450
+ "9000228_16.jpg",
451
+ "9000228_17.jpg",
452
+ "9000228_18.jpg",
453
+ "9000228_19.jpg",
454
+ "9000228_20.jpg",
455
+ "9000228_21.jpg",
456
+ "9000228_22.jpg",
457
+ "9000228_23.jpg"
458
+ ],
459
+ "conversations": [
460
+ {
461
+ "from": "human",
462
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A tranquil tableau of in the corner of the minimalist tea room, a bonsai tree added a touch of nature's beauty to the otherwise simple and elegant space\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
463
+ },
464
+ {
465
+ "from": "gpt",
466
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
467
+ }
468
+ ]
469
+ },
470
+ {
471
+ "id": "9000432",
472
+ "images": [
473
+ "9000432_00.jpg",
474
+ "9000432_01.jpg",
475
+ "9000432_02.jpg",
476
+ "9000432_03.jpg",
477
+ "9000432_04.jpg",
478
+ "9000432_05.jpg",
479
+ "9000432_06.jpg",
480
+ "9000432_07.jpg",
481
+ "9000432_08.jpg",
482
+ "9000432_09.jpg",
483
+ "9000432_10.jpg",
484
+ "9000432_11.jpg",
485
+ "9000432_12.jpg",
486
+ "9000432_13.jpg",
487
+ "9000432_14.jpg",
488
+ "9000432_15.jpg",
489
+ "9000432_16.jpg",
490
+ "9000432_17.jpg",
491
+ "9000432_18.jpg",
492
+ "9000432_19.jpg",
493
+ "9000432_20.jpg",
494
+ "9000432_21.jpg",
495
+ "9000432_22.jpg",
496
+ "9000432_23.jpg"
497
+ ],
498
+ "conversations": [
499
+ {
500
+ "from": "human",
501
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a cat drinking water\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
502
+ },
503
+ {
504
+ "from": "gpt",
505
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 2\n"
506
+ }
507
+ ]
508
+ },
509
+ {
510
+ "id": "9000095",
511
+ "images": [
512
+ "9000095_00.jpg",
513
+ "9000095_01.jpg",
514
+ "9000095_02.jpg",
515
+ "9000095_03.jpg",
516
+ "9000095_04.jpg",
517
+ "9000095_05.jpg",
518
+ "9000095_06.jpg",
519
+ "9000095_07.jpg",
520
+ "9000095_08.jpg",
521
+ "9000095_09.jpg",
522
+ "9000095_10.jpg",
523
+ "9000095_11.jpg",
524
+ "9000095_12.jpg",
525
+ "9000095_13.jpg",
526
+ "9000095_14.jpg",
527
+ "9000095_15.jpg",
528
+ "9000095_16.jpg",
529
+ "9000095_17.jpg",
530
+ "9000095_18.jpg",
531
+ "9000095_19.jpg",
532
+ "9000095_20.jpg",
533
+ "9000095_21.jpg",
534
+ "9000095_22.jpg",
535
+ "9000095_23.jpg"
536
+ ],
537
+ "conversations": [
538
+ {
539
+ "from": "human",
540
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"The bund Shanghai, watercolor painting\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
541
+ },
542
+ {
543
+ "from": "gpt",
544
+ "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
545
+ }
546
+ ]
547
+ },
548
+ {
549
+ "id": "9000558",
550
+ "images": [
551
+ "9000558_00.jpg",
552
+ "9000558_01.jpg",
553
+ "9000558_02.jpg",
554
+ "9000558_03.jpg",
555
+ "9000558_04.jpg",
556
+ "9000558_05.jpg",
557
+ "9000558_06.jpg",
558
+ "9000558_07.jpg",
559
+ "9000558_08.jpg",
560
+ "9000558_09.jpg",
561
+ "9000558_10.jpg",
562
+ "9000558_11.jpg",
563
+ "9000558_12.jpg",
564
+ "9000558_13.jpg",
565
+ "9000558_14.jpg",
566
+ "9000558_15.jpg",
567
+ "9000558_16.jpg",
568
+ "9000558_17.jpg",
569
+ "9000558_18.jpg",
570
+ "9000558_19.jpg",
571
+ "9000558_20.jpg",
572
+ "9000558_21.jpg",
573
+ "9000558_22.jpg",
574
+ "9000558_23.jpg"
575
+ ],
576
+ "conversations": [
577
+ {
578
+ "from": "human",
579
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a pink bowl\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
580
+ },
581
+ {
582
+ "from": "gpt",
583
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n"
584
+ }
585
+ ]
586
+ },
587
+ {
588
+ "id": "9000759",
589
+ "images": [
590
+ "9000759_00.jpg",
591
+ "9000759_01.jpg",
592
+ "9000759_02.jpg",
593
+ "9000759_03.jpg",
594
+ "9000759_04.jpg",
595
+ "9000759_05.jpg",
596
+ "9000759_06.jpg",
597
+ "9000759_07.jpg",
598
+ "9000759_08.jpg",
599
+ "9000759_09.jpg",
600
+ "9000759_10.jpg",
601
+ "9000759_11.jpg",
602
+ "9000759_12.jpg",
603
+ "9000759_13.jpg",
604
+ "9000759_14.jpg",
605
+ "9000759_15.jpg",
606
+ "9000759_16.jpg",
607
+ "9000759_17.jpg",
608
+ "9000759_18.jpg",
609
+ "9000759_19.jpg",
610
+ "9000759_20.jpg",
611
+ "9000759_21.jpg",
612
+ "9000759_22.jpg",
613
+ "9000759_23.jpg"
614
+ ],
615
+ "conversations": [
616
+ {
617
+ "from": "human",
618
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A fantasy landscape\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
619
+ },
620
+ {
621
+ "from": "gpt",
622
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n"
623
+ }
624
+ ]
625
+ },
626
+ {
627
+ "id": "9000692",
628
+ "images": [
629
+ "9000692_00.jpg",
630
+ "9000692_01.jpg",
631
+ "9000692_02.jpg",
632
+ "9000692_03.jpg",
633
+ "9000692_04.jpg",
634
+ "9000692_05.jpg",
635
+ "9000692_06.jpg",
636
+ "9000692_07.jpg",
637
+ "9000692_08.jpg",
638
+ "9000692_09.jpg",
639
+ "9000692_10.jpg",
640
+ "9000692_11.jpg",
641
+ "9000692_12.jpg",
642
+ "9000692_13.jpg",
643
+ "9000692_14.jpg",
644
+ "9000692_15.jpg",
645
+ "9000692_16.jpg",
646
+ "9000692_17.jpg",
647
+ "9000692_18.jpg",
648
+ "9000692_19.jpg",
649
+ "9000692_20.jpg",
650
+ "9000692_21.jpg",
651
+ "9000692_22.jpg",
652
+ "9000692_23.jpg"
653
+ ],
654
+ "conversations": [
655
+ {
656
+ "from": "human",
657
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"Snow rocky mountains peaks canyon\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
658
+ },
659
+ {
660
+ "from": "gpt",
661
+ "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 3\n"
662
+ }
663
+ ]
664
+ },
665
+ {
666
+ "id": "9000032",
667
+ "images": [
668
+ "9000032_00.jpg",
669
+ "9000032_01.jpg",
670
+ "9000032_02.jpg",
671
+ "9000032_03.jpg",
672
+ "9000032_04.jpg",
673
+ "9000032_05.jpg",
674
+ "9000032_06.jpg",
675
+ "9000032_07.jpg",
676
+ "9000032_08.jpg",
677
+ "9000032_09.jpg",
678
+ "9000032_10.jpg",
679
+ "9000032_11.jpg",
680
+ "9000032_12.jpg",
681
+ "9000032_13.jpg",
682
+ "9000032_14.jpg",
683
+ "9000032_15.jpg",
684
+ "9000032_16.jpg",
685
+ "9000032_17.jpg",
686
+ "9000032_18.jpg",
687
+ "9000032_19.jpg",
688
+ "9000032_20.jpg",
689
+ "9000032_21.jpg",
690
+ "9000032_22.jpg",
691
+ "9000032_23.jpg"
692
+ ],
693
+ "conversations": [
694
+ {
695
+ "from": "human",
696
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A person is sweeping floor\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
697
+ },
698
+ {
699
+ "from": "gpt",
700
+ "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
701
+ }
702
+ ]
703
+ },
704
+ {
705
+ "id": "9000114",
706
+ "images": [
707
+ "9000114_00.jpg",
708
+ "9000114_01.jpg",
709
+ "9000114_02.jpg",
710
+ "9000114_03.jpg",
711
+ "9000114_04.jpg",
712
+ "9000114_05.jpg",
713
+ "9000114_06.jpg",
714
+ "9000114_07.jpg",
715
+ "9000114_08.jpg",
716
+ "9000114_09.jpg",
717
+ "9000114_10.jpg",
718
+ "9000114_11.jpg",
719
+ "9000114_12.jpg",
720
+ "9000114_13.jpg",
721
+ "9000114_14.jpg",
722
+ "9000114_15.jpg",
723
+ "9000114_16.jpg",
724
+ "9000114_17.jpg",
725
+ "9000114_18.jpg",
726
+ "9000114_19.jpg",
727
+ "9000114_20.jpg",
728
+ "9000114_21.jpg",
729
+ "9000114_22.jpg",
730
+ "9000114_23.jpg"
731
+ ],
732
+ "conversations": [
733
+ {
734
+ "from": "human",
735
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"A couple in formal evening wear going home get caught in a heavy downpour with umbrellas, tilt down\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
736
+ },
737
+ {
738
+ "from": "gpt",
739
+ "value": "visual quality: 2\n temporal consistency: 1\n dynamic degree: 3\n text-to-video alignment: 3\n factual consistency: 1\n"
740
+ }
741
+ ]
742
+ },
743
+ {
744
+ "id": "9000250",
745
+ "images": [
746
+ "9000250_00.jpg",
747
+ "9000250_01.jpg",
748
+ "9000250_02.jpg",
749
+ "9000250_03.jpg",
750
+ "9000250_04.jpg",
751
+ "9000250_05.jpg",
752
+ "9000250_06.jpg",
753
+ "9000250_07.jpg",
754
+ "9000250_08.jpg",
755
+ "9000250_09.jpg",
756
+ "9000250_10.jpg",
757
+ "9000250_11.jpg",
758
+ "9000250_12.jpg",
759
+ "9000250_13.jpg",
760
+ "9000250_14.jpg",
761
+ "9000250_15.jpg",
762
+ "9000250_16.jpg",
763
+ "9000250_17.jpg",
764
+ "9000250_18.jpg",
765
+ "9000250_19.jpg",
766
+ "9000250_20.jpg",
767
+ "9000250_21.jpg",
768
+ "9000250_22.jpg",
769
+ "9000250_23.jpg"
770
+ ],
771
+ "conversations": [
772
+ {
773
+ "from": "human",
774
+ "value": "\nSuppose you are an expert in judging and evaluating the quality of AI-generated videos, \nplease watch the following frames of a given video and see the text prompt for generating the video, \nthen give scores from 5 different dimensions:\n(1) visual quality: the quality of the video in terms of clearness, resolution, brightness, and color\n(2) temporal consistency, the consistency of objects or humans in video\n(3) dynamic degree, the degree of dynamic changes\n(4) text-to-video alignment, the alignment between the text prompt and the video content\n(5) factual consistency, the consistency of the video content with the common-sense and factual knowledge\n\nFor each dimension, output a number from [1,2,3,4], \nin which '1' means 'Bad', '2' means 'Average', '3' means 'Good', \n'4' means 'Real' or 'Perfect' (the video is like a real video)\nHere is an output example:\nvisual quality: 4\ntemporal consistency: 4\ndynamic degree: 3\ntext-to-video alignment: 1\nfactual consistency: 2\n\nFor this video, the text prompt is \"a toothbrush and a sink\",\nall the frames of video are as follows: \n\n<image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> <image> \n"
775
+ },
776
+ {
777
+ "from": "gpt",
778
+ "value": "visual quality: 4\n temporal consistency: 4\n dynamic degree: 4\n text-to-video alignment: 4\n factual consistency: 4\n"
779
+ }
780
+ ]
781
+ }
782
+ ]