upload refined_40k/test_bench 's json, videos and frames

data/refined_40k/test_bench/data_bad_img.json

@@ -0,0 +1,392 @@
+[
+    {
+        "id": "g000072",
+        "images": [
+            "g000072_00.jpg",
+            "g000072_01.jpg",
+            "g000072_02.jpg",
+            "g000072_03.jpg",
+            "g000072_04.jpg",
+            "g000072_05.jpg",
+            "g000072_06.jpg",
+            "g000072_07.jpg",
+            "g000072_08.jpg",
+            "g000072_09.jpg",
+            "g000072_10.jpg",
+            "g000072_11.jpg",
+            "g000072_12.jpg",
+            "g000072_13.jpg",
+            "g000072_14.jpg",
+            "g000072_15.jpg",
+            "g000072_16.jpg",
+            "g000072_17.jpg",
+            "g000072_18.jpg",
+            "g000072_19.jpg",
+            "g000072_20.jpg",
+            "g000072_21.jpg",
+            "g000072_22.jpg",
+            "g000072_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"boogie nights performing in outer space with aliens  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 3\n"
+            }
+        ]
+    },
+    {
+        "id": "g000732",
+        "images": [
+            "g000732_00.jpg",
+            "g000732_01.jpg",
+            "g000732_02.jpg",
+            "g000732_03.jpg",
+            "g000732_04.jpg",
+            "g000732_05.jpg",
+            "g000732_06.jpg",
+            "g000732_07.jpg",
+            "g000732_08.jpg",
+            "g000732_09.jpg",
+            "g000732_10.jpg",
+            "g000732_11.jpg",
+            "g000732_12.jpg",
+            "g000732_13.jpg",
+            "g000732_14.jpg",
+            "g000732_15.jpg",
+            "g000732_16.jpg",
+            "g000732_17.jpg",
+            "g000732_18.jpg",
+            "g000732_19.jpg",
+            "g000732_20.jpg",
+            "g000732_21.jpg",
+            "g000732_22.jpg",
+            "g000732_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 baby walks across a room to a black bag.\",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000600",
+        "images": [
+            "g000600_00.jpg",
+            "g000600_01.jpg",
+            "g000600_02.jpg",
+            "g000600_03.jpg",
+            "g000600_04.jpg",
+            "g000600_05.jpg",
+            "g000600_06.jpg",
+            "g000600_07.jpg",
+            "g000600_08.jpg",
+            "g000600_09.jpg",
+            "g000600_10.jpg",
+            "g000600_11.jpg",
+            "g000600_12.jpg",
+            "g000600_13.jpg",
+            "g000600_14.jpg",
+            "g000600_15.jpg",
+            "g000600_16.jpg",
+            "g000600_17.jpg",
+            "g000600_18.jpg",
+            "g000600_19.jpg",
+            "g000600_20.jpg",
+            "g000600_21.jpg",
+            "g000600_22.jpg",
+            "g000600_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 soccer match is taking place in an outdoor stadium with a large screen displaying the scoreboard.\",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000197",
+        "images": [
+            "g000197_00.jpg",
+            "g000197_01.jpg",
+            "g000197_02.jpg",
+            "g000197_03.jpg",
+            "g000197_04.jpg",
+            "g000197_05.jpg",
+            "g000197_06.jpg",
+            "g000197_07.jpg",
+            "g000197_08.jpg",
+            "g000197_09.jpg",
+            "g000197_10.jpg",
+            "g000197_11.jpg",
+            "g000197_12.jpg",
+            "g000197_13.jpg",
+            "g000197_14.jpg",
+            "g000197_15.jpg",
+            "g000197_16.jpg",
+            "g000197_17.jpg",
+            "g000197_18.jpg",
+            "g000197_19.jpg",
+            "g000197_20.jpg",
+            "g000197_21.jpg",
+            "g000197_22.jpg",
+            "g000197_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 man is standing in front of a window with his shirt off, showing his muscular chest. he is wearing blue jeans and a white shirt.\",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000047",
+        "images": [
+            "g000047_00.jpg",
+            "g000047_01.jpg",
+            "g000047_02.jpg",
+            "g000047_03.jpg",
+            "g000047_04.jpg",
+            "g000047_05.jpg",
+            "g000047_06.jpg",
+            "g000047_07.jpg",
+            "g000047_08.jpg",
+            "g000047_09.jpg",
+            "g000047_10.jpg",
+            "g000047_11.jpg",
+            "g000047_12.jpg",
+            "g000047_13.jpg",
+            "g000047_14.jpg",
+            "g000047_15.jpg",
+            "g000047_16.jpg",
+            "g000047_17.jpg",
+            "g000047_18.jpg",
+            "g000047_19.jpg",
+            "g000047_20.jpg",
+            "g000047_21.jpg",
+            "g000047_22.jpg",
+            "g000047_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"movie advertisement style, long beard grows longer on young handsome man  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 3\n"
+            }
+        ]
+    },
+    {
+        "id": "g000235",
+        "images": [
+            "g000235_00.jpg",
+            "g000235_01.jpg",
+            "g000235_02.jpg",
+            "g000235_03.jpg",
+            "g000235_04.jpg",
+            "g000235_05.jpg",
+            "g000235_06.jpg",
+            "g000235_07.jpg",
+            "g000235_08.jpg",
+            "g000235_09.jpg",
+            "g000235_10.jpg",
+            "g000235_11.jpg",
+            "g000235_12.jpg",
+            "g000235_13.jpg",
+            "g000235_14.jpg",
+            "g000235_15.jpg",
+            "g000235_16.jpg",
+            "g000235_17.jpg",
+            "g000235_18.jpg",
+            "g000235_19.jpg",
+            "g000235_20.jpg",
+            "g000235_21.jpg",
+            "g000235_22.jpg",
+            "g000235_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"maa durga and lord shiva doing shopping in AI world  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 3\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 3\n"
+            }
+        ]
+    },
+    {
+        "id": "g000894",
+        "images": [
+            "g000894_00.jpg",
+            "g000894_01.jpg",
+            "g000894_02.jpg",
+            "g000894_03.jpg",
+            "g000894_04.jpg",
+            "g000894_05.jpg",
+            "g000894_06.jpg",
+            "g000894_07.jpg",
+            "g000894_08.jpg",
+            "g000894_09.jpg",
+            "g000894_10.jpg",
+            "g000894_11.jpg",
+            "g000894_12.jpg",
+            "g000894_13.jpg",
+            "g000894_14.jpg",
+            "g000894_15.jpg",
+            "g000894_16.jpg",
+            "g000894_17.jpg",
+            "g000894_18.jpg",
+            "g000894_19.jpg",
+            "g000894_20.jpg",
+            "g000894_21.jpg",
+            "g000894_22.jpg",
+            "g000894_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 beach, a swimming girl  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000688",
+        "images": [
+            "g000688_00.jpg",
+            "g000688_01.jpg",
+            "g000688_02.jpg",
+            "g000688_03.jpg",
+            "g000688_04.jpg",
+            "g000688_05.jpg",
+            "g000688_06.jpg",
+            "g000688_07.jpg",
+            "g000688_08.jpg",
+            "g000688_09.jpg",
+            "g000688_10.jpg",
+            "g000688_11.jpg",
+            "g000688_12.jpg",
+            "g000688_13.jpg",
+            "g000688_14.jpg",
+            "g000688_15.jpg",
+            "g000688_16.jpg",
+            "g000688_17.jpg",
+            "g000688_18.jpg",
+            "g000688_19.jpg",
+            "g000688_20.jpg",
+            "g000688_21.jpg",
+            "g000688_22.jpg",
+            "g000688_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"exploration of level 0 in the backrooms  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000918",
+        "images": [
+            "g000918_00.jpg",
+            "g000918_01.jpg",
+            "g000918_02.jpg",
+            "g000918_03.jpg",
+            "g000918_04.jpg",
+            "g000918_05.jpg",
+            "g000918_06.jpg",
+            "g000918_07.jpg",
+            "g000918_08.jpg",
+            "g000918_09.jpg",
+            "g000918_10.jpg",
+            "g000918_11.jpg",
+            "g000918_12.jpg",
+            "g000918_13.jpg",
+            "g000918_14.jpg",
+            "g000918_15.jpg",
+            "g000918_16.jpg",
+            "g000918_17.jpg",
+            "g000918_18.jpg",
+            "g000918_19.jpg",
+            "g000918_20.jpg",
+            "g000918_21.jpg",
+            "g000918_22.jpg",
+            "g000918_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"Illustrate a scene where Kaito helps the trapped colorful bird escape from the thorny vines in the garden.  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 2\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    },
+    {
+        "id": "g000375",
+        "images": [
+            "g000375_00.jpg",
+            "g000375_01.jpg",
+            "g000375_02.jpg",
+            "g000375_03.jpg",
+            "g000375_04.jpg",
+            "g000375_05.jpg",
+            "g000375_06.jpg",
+            "g000375_07.jpg",
+            "g000375_08.jpg",
+            "g000375_09.jpg",
+            "g000375_10.jpg",
+            "g000375_11.jpg",
+            "g000375_12.jpg",
+            "g000375_13.jpg",
+            "g000375_14.jpg",
+            "g000375_15.jpg",
+            "g000375_16.jpg",
+            "g000375_17.jpg",
+            "g000375_18.jpg",
+            "g000375_19.jpg",
+            "g000375_20.jpg",
+            "g000375_21.jpg",
+            "g000375_22.jpg",
+            "g000375_23.jpg"
+        ],
+        "conversations": [
+            {
+                "from": "human",
+                "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 \"drone shot of a a town being rained down upon by comets and meteors  \",\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"
+            },
+            {
+                "from": "gpt",
+                "value": "visual quality: 1\n temporal consistency: 3\n dynamic degree: 1\n text-to-video alignment: 3\n factual consistency: 1\n"
+            }
+        ]
+    }
+]