ServiceNow/coco_encoded
收藏资源简介:
--- dataset_info: features: - name: image dtype: image - name: filepath dtype: string - name: sentids list: int32 - name: filename dtype: string - name: imgid dtype: int32 - name: split dtype: string - name: sentences struct: - name: tokens list: string - name: raw dtype: string - name: imgid dtype: int32 - name: sentid dtype: int32 - name: cocoid dtype: int32 - name: hf_hub_laion/CLIP_ViT_g_14_laion2B_s12B_b42K_short_features sequence: float32 - name: FacebookAI/roberta_base_short_features sequence: float32 - name: FacebookAI/roberta_base_long_features sequence: float32 - name: FacebookAI/roberta_base_normalized_long_features sequence: float32 - name: facebook/dinov2_large_short_features sequence: float32 - name: facebook/dinov2_large_long_features sequence: float32 - name: facebook/dinov2_large_normalized_long_features sequence: float32 - name: hf_hub_timm/ViT_B_16_SigLIP_short_features sequence: float32 splits: - name: train num_bytes: 102385311908.625 num_examples: 566747 - name: validation num_bytes: 4517702812.75 num_examples: 25010 - name: test num_bytes: 4516243134.75 num_examples: 25010 download_size: 80421808435 dataset_size: 111419257856.125 configs: - config_name: default data_files: - split: train path: data/train-* - split: validation path: data/validation-* - split: test path: data/test-* ---
数据集信息: ### 特征字段 本数据集涵盖如下特征字段: 1. `image`:图像类型数据 2. `filepath`:字符串格式的文件路径 3. `sentids`:int32类型的句子ID列表 4. `filename`:字符串格式的文件名 5. `imgid`:int32类型的图像ID 6. `split`:字符串格式的数据集划分标识 7. `sentences`:结构体类型特征,包含子字段: - `tokens`:字符串类型的标记(Token)列表 - `raw`:字符串类型的原始句子文本 - `imgid`:int32类型的关联图像ID - `sentid`:int32类型的句子ID 8. `cocoid`:int32类型的COCO图像ID 9. `hf_hub_laion/CLIP_ViT_g_14_laion2B_s12B_b42K_short_features`:float32类型的特征序列 10. `FacebookAI/roberta_base_short_features`:float32类型的特征序列 11. `FacebookAI/roberta_base_long_features`:float32类型的特征序列 12. `FacebookAI/roberta_base_normalized_long_features`:float32类型的归一化长文本特征序列 13. `facebook/dinov2_large_short_features`:float32类型的特征序列 14. `facebook/dinov2_large_long_features`:float32类型的特征序列 15. `facebook/dinov2_large_normalized_long_features`:float32类型的归一化长文本特征序列 16. `hf_hub_timm/ViT_B_16_SigLIP_short_features`:float32类型的特征序列 ### 数据集划分 本数据集包含三类划分子集: - 训练集(train):数据体量为102385311908.625字节,样本量566747 - 验证集(validation):数据体量为4517702812.75字节,样本量25010 - 测试集(test):数据体量为4516243134.75字节,样本量25010 本数据集整体下载规模为80421808435字节,总存储体量为111419257856.125字节。 ### 数据集配置 默认配置对应的数据文件路径如下: - 训练集:`data/train-*` - 验证集:`data/validation-*` - 测试集:`data/test-*`
数据集概述
数据特征
- image: 图像数据
- filepath: 字符串,文件路径
- sentids: 整数列表
- filename: 字符串,文件名
- imgid: 整数,图像ID
- split: 字符串,数据集划分(如训练、验证、测试)
- sentences: 结构化数据,包含以下字段:
- tokens: 字符串列表
- raw: 字符串,原始文本
- imgid: 整数,图像ID
- sentid: 整数,句子ID
- cocoid: 整数,COCO图像ID
- hf_hub_laion/CLIP_ViT_g_14_laion2B_s12B_b42K_short_features: 浮点数序列
- FacebookAI/roberta_base_short_features: 浮点数序列
- FacebookAI/roberta_base_long_features: 浮点数序列
- FacebookAI/roberta_base_normalized_long_features: 浮点数序列
- facebook/dinov2_large_short_features: 浮点数序列
- facebook/dinov2_large_long_features: 浮点数序列
- facebook/dinov2_large_normalized_long_features: 浮点数序列
- hf_hub_timm/ViT_B_16_SigLIP_short_features: 浮点数序列
数据集划分
- train: 训练集,包含566747个样本,大小为102385311908.625字节
- validation: 验证集,包含25010个样本,大小为4517702812.75字节
- test: 测试集,包含25010个样本,大小为4516243134.75字节
数据集大小
- 下载大小: 80421808435字节
- 数据集大小: 111419257856.125字节
配置
- config_name: default
- data_files:
- train: 路径为
data/train-* - validation: 路径为
data/validation-* - test: 路径为
data/test-*
- train: 路径为
- data_files:



