AnchorSR/TrainingData_Stage3_VideoCases

Stage3 视频／多图人工查看样例

这是 18条人工查看样例，不是新训练集或评测集。直接向下浏览 QA 和拼图， 也可在 Dataset Viewer 中查看 frames 图片列。训练 QA 与原数据逐字保留；答案是数据集标注，不是人工确认的视觉真值。

来源	已确认视频类	待判断多图
CA-VQA	3（有序帧）	0
SpaceVista	3（有序帧）	3
VSI-590K	3（视频文件）	0
SIMS-VSI	3（视频文件）	0
ViCA-322K	3（视频文件）	0

选样：从正式 Small/train 中按固定 ID 顺序选择，优先不同任务、不同来源场景；不是随机代表性统计，也没有按答案是否正确挑样。 每张拼图从左到右、从上到下阅读。视频文件全程均匀抽取最多16帧；帧序列保留全部输入帧。 展示缩放不改变正式训练数据；拼图不是模型训练输入。多图未知项不计为视频。

来源：AnchorSR/TrainingData_Stage3， 固定提交 a6cb55497a991b327b32410d946e41d3bf0f3610，Small/train（同时属于 Large）。 每条提供 sample_id、来源版本、场景、原始 QA、实际训练 QA 与输入分组。

许可沿用原始数据，此仓库不重新授权：CA-VQA/CubifyAnything 为 CC-BY-NC-ND-4.0； SpaceVista 为 CC-BY-4.0；VSI-590K、SIMS-VSI 为 Apache-2.0；ViCA-322K 为 CC-BY-NC-4.0。 上游媒体自身的许可条件仍适用。不得将混合集视为可自由商用或重新许可的素材。 本仓库仅供检查；需遵守源数据对使用、修改和再分发的限制。

01 · cavqa · camera_depth

模态：video_frame_sequence；许可：cc-by-nc-nd-4.0。

Q（训练输入）

How far away is the painting?

A（训练答案）

1.48 m

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 REF是末尾参考帧：前面的支持帧加参考帧不应理解为严格连续播放的视频。

原始 QA 与样例 ID

{
  "question": "How far away is the painting?",
  "raw_answer": "1.48m"
}

cavqa:00000-of-01024:100:99

02 · cavqa · height

模态：video_frame_sequence；许可：cc-by-nc-nd-4.0。

Q（训练输入）

How tall is the book?

A（训练答案）

9 cm

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 REF是末尾参考帧：前面的支持帧加参考帧不应理解为严格连续播放的视频。

原始 QA 与样例 ID

{
  "question": "How tall is the book?",
  "raw_answer": "9cm"
}

cavqa:00000-of-01024:102:28

03 · cavqa · length

模态：video_frame_sequence；许可：cc-by-nc-nd-4.0。

Q（训练输入）

What is the length of vacuum cleaner?

A（训练答案）

32 cm

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 REF是末尾参考帧：前面的支持帧加参考帧不应理解为严格连续播放的视频。

原始 QA 与样例 ID

{
  "question": "What is the length of vacuum cleaner?",
  "raw_answer": "32cm"
}

cavqa:00000-of-01024:111:20

04 · spacevista · area

模态：video_frame_sequence；许可：cc-by-4.0。

Q（训练输入）

Estimate the floor space of the room visible in the footage (in square meters).

A（训练答案）

20 m2

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。

原始 QA 与样例 ID

{
  "question": "Estimate the floor space of the room visible in the footage (in square meters).",
  "raw_answer": "20"
}

spacevista:1001419

05 · spacevista · camera_angle

模态：video_frame_sequence；许可：cc-by-4.0。

Q（训练输入）

Please estimate the overall rotation of the camera in the video, ignoring translation.
Report the angle in degrees.

A（训练答案）

53 deg

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。

原始 QA 与样例 ID

{
  "question": "Please estimate the overall rotation of the camera in the video, ignoring translation.",
  "raw_answer": "53"
}

spacevista:1000152

06 · spacevista · camera_depth

模态：video_frame_sequence；许可：cc-by-4.0。

Q（训练输入）

Provide the range to the object that the red box frames in the first video frame (in centimeters).

A（训练答案）

12.7 cm

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。

原始 QA 与样例 ID

{
  "question": "Provide the range to the object that the red box frames in the first video frame (in centimeters).",
  "raw_answer": "12.7"
}

spacevista:1000004

07 · vsi590k · angle

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
At the door, facing the chair, what's the precise angle of clockwise rotation required to turn toward the trash bin?
Please answer the question using a single word or phrase.
Report the angle in degrees.

A（训练答案）

18 deg

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nThese are frames of a video.\nAt the door, facing the chair, what's the precise angle of clockwise rotation required to turn toward the trash bin?\nPlease answer the question using a single word or phrase.",
  "raw_answer": "18"
}

vsi590k:10

08 · vsi590k · area

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
Please indicate the size of the room using square feet. If there are multiple rooms, estimate the combined area.
Please answer the question using a single word or phrase.

A（训练答案）

192 ft2

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nThese are frames of a video.\nPlease indicate the size of the room using square feet. If there are multiple rooms, estimate the combined area.\nPlease answer the question using a single word or phrase.",
  "raw_answer": "192"
}

vsi590k:178066

09 · vsi590k · closest_surface

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
Specify precisely how far apart the whiteboard and the telephone are at their closest points, expressed in centimeters.
Please answer the question using a single word or phrase.

A（训练答案）

240 cm

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nThese are frames of a video.\nSpecify precisely how far apart the whiteboard and the telephone are at their closest points, expressed in centimeters.\nPlease answer the question using a single word or phrase.",
  "raw_answer": "240.0"
}

vsi590k:155095

10 · sims_vsi · area

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
What is the size of this room (in square meters)?
If multiple rooms are shown, estimate the size of the combined space.
Please answer the question using a single word or phrase.

A（训练答案）

263.7 m2

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>These are frames of a video.\nWhat is the size of this room (in square meters)?\nIf multiple rooms are shown, estimate the size of the combined space.\nPlease answer the question using a single word or phrase.",
  "raw_answer": "263.7"
}

sims_vsi:qas/vsi_room_size_est_oe_seed_0.jsonl:0

11 · sims_vsi · longest_side

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
What is the length of the longest dimension (length, width, or height) of the painting, measured in centimeters?
Please answer the question using a single word or phrase.

A（训练答案）

80 cm

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>These are frames of a video.\nWhat is the length of the longest dimension (length, width, or height) of the painting, measured in centimeters?\nPlease answer the question using a single word or phrase.",
  "raw_answer": "80"
}

sims_vsi:qas/vsi_obj_size_est_oe_seed_0.jsonl:0

12 · sims_vsi · surface_distance

模态：video_file；许可：apache-2.0。

Q（训练输入）

These are frames of a video.
Measuring from the closest point of each object, what is the direct distance between the clothes dryer and the handcart (in meters)?
Please answer the question using a single word or phrase.

A（训练答案）

8.3 m

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>These are frames of a video.\nMeasuring from the closest point of each object, what is the direct distance between the clothes dryer and the handcart (in meters)?\nPlease answer the question using a single word or phrase.",
  "raw_answer": "8.3"
}

sims_vsi:qas/vsi_obj_abs_distance_oe_seed_0.jsonl:10

13 · vica322k · area

模态：video_file；许可：cc-by-nc-4.0。

Q（训练输入）

Determine the total area of this room in square meters. If multiple rooms are present, estimate the combined space.

A（训练答案）

12.72 m2

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nDetermine the total area of this room in square meters. If multiple rooms are present, estimate the combined space.",
  "raw_answer": "12.72"
}

vica322k:arkitscenes/base/room_size.json:0

14 · vica322k · longest_side

模态：video_file；许可：cc-by-nc-4.0。

Q（训练输入）

How long is the largest side (length, width, or height) of the sink in centimeters?

A（训练答案）

78 cm

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nHow long is the largest side (length, width, or height) of the sink in centimeters?",
  "raw_answer": "78"
}

vica322k:arkitscenes/base/object_size_estimation.json:0

15 · vica322k · surface_distance

模态：video_file；许可：cc-by-nc-4.0。

Q（训练输入）

How many meters apart are the closest points of sink and toilet?

A（训练答案）

1.7 m

MP4 视频按完整解码帧范围均匀抽取最多16帧；仅用于展示，不代表训练抽帧策略。帧号从0开始。

原始 QA 与样例 ID

{
  "question": "<image>\nHow many meters apart are the closest points of sink and toilet?",
  "raw_answer": "1.7"
}

vica322k:arkitscenes/base/object_abs_distance.json:1

16 · spacevista · area

模态：multi_image_unresolved；许可：cc-by-4.0。

Q（训练输入）

From the video, calculate the approximate total size of the room in square meters. Your input must be limited to a single number.
Please give your final answer between the <answer> </answer> tags.

A（训练答案）

<answer>10 m2</answer>

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 此例尚不能确认时间顺序或视频属性，请人工判断。

原始 QA 与样例 ID

{
  "question": "From the video, calculate the approximate total size of the room in square meters. Your input must be limited to a single number.\nPlease give your final answer between the <answer> </answer> tags.",
  "raw_answer": "10"
}

spacevista:1001459

17 · spacevista · camera_depth

模态：multi_image_unresolved；许可：cc-by-4.0。

Q（训练输入）

If the center of the paper towel roll (red point) is 1.5 meters deep, what is the depth of trash bin (blue point)?  Calculate or judge based on the 3D center points of these objects. Input a single number to complete your answer.
Please give your final answer between the <answer> </answer> tags.

A（训练答案）

<answer>1.2 m</answer>

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 此例尚不能确认时间顺序或视频属性，请人工判断。

原始 QA 与样例 ID

{
  "question": "If the center of the paper towel roll (red point) is 1.5 meters deep, what is the depth of trash bin (blue point)?  Calculate or judge based on the 3D center points of these objects. Input a single number to complete your answer.\nPlease give your final answer between the <answer> </answer> tags.",
  "raw_answer": "1.2"
}

spacevista:1155590

18 · spacevista · center_depth

模态：multi_image_unresolved；许可：cc-by-4.0。

Q（训练输入）

The depth of sofa chair (red point) is given as 3.1. Calculate how far apart in depth table (green point) and column (blue point) are from each other in meters. Calculate or judge based on the 3D center points of these objects. The depth is calculated based on the image where the markers corresponding to these objects are located.
Provide your reasoning in <think> </think>.
Provide only the numerical value (e.g., 42 or 3.14) in meters in <answer> </answer>.

A（训练答案）

<answer>0.5</answer>

全部输入图片按原分组和顺序展示，保留重复帧及读取器渲染的框／点／掩码；G表示源分组，input编号从1开始，不是原视频帧号。 此例尚不能确认时间顺序或视频属性，请人工判断。

原始 QA 与样例 ID

{
  "question": "The depth of sofa chair (red point) is given as 3.1. Calculate how far apart in depth table (green point) and column (blue point) are from each other in meters. Calculate or judge based on the 3D center points of these objects. The depth is calculated based on the image where the markers corresponding to these objects are located.\nProvide your reasoning in <think> </think>.\nProvide only the numerical value (e.g., 42 or 3.14) in meters in <answer> </answer>.",
  "raw_answer": "0.5"
}

spacevista:1000201