Name: AbstractPhil/bulk-coco-features
Creator: AbstractPhil
Published: 2025-12-28 22:13:59
License: 暂无描述

下载链接：

https://hf-mirror.com/datasets/AbstractPhil/bulk-coco-features

下载链接

链接失效反馈

官方服务：

资源简介：

--- license: cc-by-4.0 task_categories: - feature-extraction dataset_info: - config_name: clip_b16_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 512 - name: labels list: int32 splits: - name: train num_bytes: 245043204 num_examples: 118287 - name: val num_bytes: 10358524 num_examples: 5000 download_size: 130923096 dataset_size: 255401728 - config_name: clip_b16_openai features: - name: image_id dtype: int64 - name: features list: float32 length: 512 - name: labels list: int32 splits: - name: train num_bytes: 245043204 num_examples: 118287 - name: val num_bytes: 10358524 num_examples: 5000 download_size: 130944684 dataset_size: 255401728 - config_name: clip_b32_datacomp features: - name: image_id dtype: int64 - name: features list: float32 length: 512 - name: labels list: int32 splits: - name: train num_bytes: 245043204 num_examples: 118287 - name: val num_bytes: 10358524 num_examples: 5000 download_size: 130947622 dataset_size: 255401728 - config_name: clip_b32_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 512 - name: labels list: int32 splits: - name: train num_bytes: 245043204 num_examples: 118287 - name: val num_bytes: 10358524 num_examples: 5000 download_size: 131076723 dataset_size: 255401728 - config_name: clip_b32_openai features: - name: image_id dtype: int64 - name: features list: float32 length: 512 - name: labels list: int32 splits: - name: train num_bytes: 245043204 num_examples: 118287 - name: val num_bytes: 10358524 num_examples: 5000 download_size: 130914056 dataset_size: 255401728 - config_name: clip_bigg14_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 1280 - name: labels list: int32 splits: - name: train num_bytes: 608420868 num_examples: 118287 - name: val num_bytes: 25718524 num_examples: 5000 download_size: 309633647 dataset_size: 634139392 - config_name: clip_g14_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 250761385 dataset_size: 507893504 - config_name: clip_h14_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 250708485 dataset_size: 507893504 - config_name: clip_l14_336_openai features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 190867743 dataset_size: 381647616 - config_name: clip_l14_datacomp features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 190802519 dataset_size: 381647616 - config_name: clip_l14_laion2b features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 190784016 dataset_size: 381647616 - config_name: clip_l14_openai features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 190876738 dataset_size: 381647616 - config_name: dinov2_b14 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 457255828 dataset_size: 381647616 - config_name: dinov2_b14_reg features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 457258293 dataset_size: 381647616 - config_name: dinov2_g14 features: - name: image_id dtype: int64 - name: features list: float32 length: 1536 - name: labels list: int32 splits: - name: train num_bytes: 729546756 num_examples: 118287 - name: val num_bytes: 30838524 num_examples: 5000 download_size: 836539611 dataset_size: 760385280 - config_name: dinov2_g14_reg features: - name: image_id dtype: int64 - name: features list: float32 length: 1536 - name: labels list: int32 splits: - name: train num_bytes: 729546756 num_examples: 118287 - name: val num_bytes: 30838524 num_examples: 5000 download_size: 836539678 dataset_size: 760385280 - config_name: dinov2_l14 features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 583645222 dataset_size: 507893504 - config_name: dinov2_l14_reg features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 583646758 dataset_size: 507893504 - config_name: dinov2_s14 features: - name: image_id dtype: int64 - name: features list: float32 length: 384 - name: labels list: int32 splits: - name: train num_bytes: 184480260 num_examples: 118287 - name: val num_bytes: 7798524 num_examples: 5000 download_size: 267609731 dataset_size: 192278784 - config_name: dinov2_s14_reg features: - name: image_id dtype: int64 - name: features list: float32 length: 384 - name: labels list: int32 splits: - name: train num_bytes: 184480260 num_examples: 118287 - name: val num_bytes: 7798524 num_examples: 5000 download_size: 267610384 dataset_size: 192278784 - config_name: mae_b16 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 457270563 dataset_size: 381647616 - config_name: mae_h14 features: - name: image_id dtype: int64 - name: features list: float32 length: 1280 - name: labels list: int32 splits: - name: train num_bytes: 608420868 num_examples: 118287 - name: val num_bytes: 25718524 num_examples: 5000 download_size: 710120720 dataset_size: 634139392 - config_name: mae_l16 features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 583663949 dataset_size: 507893504 - config_name: siglip2_b16_256 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 178717284 dataset_size: 381647616 - config_name: siglip2_b16_512 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 178751233 dataset_size: 381647616 - config_name: siglip2_l16_384 features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 247288440 dataset_size: 507893504 - config_name: siglip_b16_384 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 188025646 dataset_size: 381647616 - config_name: siglip_b16_512 features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 188018714 dataset_size: 381647616 - config_name: siglip_l16_256 features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 247827934 dataset_size: 507893504 - config_name: siglip_l16_384 features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 247734135 dataset_size: 507893504 - config_name: siglip_so400m_384 features: - name: image_id dtype: int64 - name: features list: float32 length: 1152 - name: labels list: int32 splits: - name: train num_bytes: 547857924 num_examples: 118287 - name: val num_bytes: 23158524 num_examples: 5000 download_size: 277493237 dataset_size: 571016448 - config_name: vit_b16_21k features: - name: image_id dtype: int64 - name: features list: float32 length: 768 - name: labels list: int32 splits: - name: train num_bytes: 366169092 num_examples: 118287 - name: val num_bytes: 15478524 num_examples: 5000 download_size: 457260731 dataset_size: 381647616 - config_name: vit_l16_21k features: - name: image_id dtype: int64 - name: features list: float32 length: 1024 - name: labels list: int32 splits: - name: train num_bytes: 487294980 num_examples: 118287 - name: val num_bytes: 20598524 num_examples: 5000 download_size: 583651349 dataset_size: 507893504 - config_name: vit_s16_21k features: - name: image_id dtype: int64 - name: features list: float32 length: 384 - name: labels list: int32 splits: - name: train num_bytes: 184480260 num_examples: 118287 - name: val num_bytes: 7798524 num_examples: 5000 download_size: 267617689 dataset_size: 192278784 configs: - config_name: clip_b16_laion2b data_files: - split: train path: clip_b16_laion2b/train-* - split: val path: clip_b16_laion2b/val-* - config_name: clip_b16_openai data_files: - split: train path: clip_b16_openai/train-* - split: val path: clip_b16_openai/val-* - config_name: clip_b32_datacomp data_files: - split: train path: clip_b32_datacomp/train-* - split: val path: clip_b32_datacomp/val-* - config_name: clip_b32_laion2b data_files: - split: train path: clip_b32_laion2b/train-* - split: val path: clip_b32_laion2b/val-* - config_name: clip_b32_openai data_files: - split: train path: clip_b32_openai/train-* - split: val path: clip_b32_openai/val-* - config_name: clip_bigg14_laion2b data_files: - split: train path: clip_bigg14_laion2b/train-* - split: val path: clip_bigg14_laion2b/val-* - config_name: clip_g14_laion2b data_files: - split: train path: clip_g14_laion2b/train-* - split: val path: clip_g14_laion2b/val-* - config_name: clip_h14_laion2b data_files: - split: train path: clip_h14_laion2b/train-* - split: val path: clip_h14_laion2b/val-* - config_name: clip_l14_336_openai data_files: - split: train path: clip_l14_336_openai/train-* - split: val path: clip_l14_336_openai/val-* - config_name: clip_l14_datacomp data_files: - split: train path: clip_l14_datacomp/train-* - split: val path: clip_l14_datacomp/val-* - config_name: clip_l14_laion2b data_files: - split: train path: clip_l14_laion2b/train-* - split: val path: clip_l14_laion2b/val-* - config_name: clip_l14_openai data_files: - split: train path: clip_l14_openai/train-* - split: val path: clip_l14_openai/val-* - config_name: dinov2_b14 data_files: - split: train path: dinov2_b14/train-* - split: val path: dinov2_b14/val-* - config_name: dinov2_b14_reg data_files: - split: train path: dinov2_b14_reg/train-* - split: val path: dinov2_b14_reg/val-* - config_name: dinov2_g14 data_files: - split: train path: dinov2_g14/train-* - split: val path: dinov2_g14/val-* - config_name: dinov2_g14_reg data_files: - split: train path: dinov2_g14_reg/train-* - split: val path: dinov2_g14_reg/val-* - config_name: dinov2_l14 data_files: - split: train path: dinov2_l14/train-* - split: val path: dinov2_l14/val-* - config_name: dinov2_l14_reg data_files: - split: train path: dinov2_l14_reg/train-* - split: val path: dinov2_l14_reg/val-* - config_name: dinov2_s14 data_files: - split: train path: dinov2_s14/train-* - split: val path: dinov2_s14/val-* - config_name: dinov2_s14_reg data_files: - split: train path: dinov2_s14_reg/train-* - split: val path: dinov2_s14_reg/val-* - config_name: mae_b16 data_files: - split: train path: mae_b16/train-* - split: val path: mae_b16/val-* - config_name: mae_h14 data_files: - split: train path: mae_h14/train-* - split: val path: mae_h14/val-* - config_name: mae_l16 data_files: - split: train path: mae_l16/train-* - split: val path: mae_l16/val-* - config_name: siglip2_b16_256 data_files: - split: train path: siglip2_b16_256/train-* - split: val path: siglip2_b16_256/val-* - config_name: siglip2_b16_512 data_files: - split: train path: siglip2_b16_512/train-* - split: val path: siglip2_b16_512/val-* - config_name: siglip2_l16_384 data_files: - split: train path: siglip2_l16_384/train-* - split: val path: siglip2_l16_384/val-* - config_name: siglip_b16_384 data_files: - split: train path: siglip_b16_384/train-* - split: val path: siglip_b16_384/val-* - config_name: siglip_b16_512 data_files: - split: train path: siglip_b16_512/train-* - split: val path: siglip_b16_512/val-* - config_name: siglip_l16_256 data_files: - split: train path: siglip_l16_256/train-* - split: val path: siglip_l16_256/val-* - config_name: siglip_l16_384 data_files: - split: train path: siglip_l16_384/train-* - split: val path: siglip_l16_384/val-* - config_name: siglip_so400m_384 data_files: - split: train path: siglip_so400m_384/train-* - split: val path: siglip_so400m_384/val-* - config_name: vit_b16_21k data_files: - split: train path: vit_b16_21k/train-* - split: val path: vit_b16_21k/val-* - config_name: vit_l16_21k data_files: - split: train path: vit_l16_21k/train-* - split: val path: vit_l16_21k/val-* - config_name: vit_s16_21k data_files: - split: train path: vit_s16_21k/train-* - split: val path: vit_s16_21k/val-* --- Here exists the bulk prepared sets for coco 2017. With this I will begin testing the first WIDE ViT-Beatrix, ViT-Zana, ViT-Beatrix-DualStream, Clip-Vit-Beatrix, GeoVit-Beans and more. These wide vits will be using new forms of formula meant to fuse structural behaviors together which exist on multiple different manifolds simultaneously. These upcoming experiments will be with established SOTA-based processes adopted and modulated for geofractal behavior from multiple transfer learning concepts. Additionally, these features will be useful to anyone who needs them.

许可协议：CC BY 4.0 任务类别：特征提取 ### 数据集详情本数据集包含多个配置项，每个配置项对应不同的预训练模型与数据集组合，各配置项均包含图像ID、特征、标签三类字段，以及训练集、验证集两种数据划分，具体配置项如下： 1. **配置名称：clip_b16_laion2b** 特征字段： - 图像ID（image_id）：64位整数类型（int64） - 特征（features）：长度为512的单精度浮点数（float32）列表 - 标签（labels）：32位整数列表数据划分： - 训练集：字节数245043204，样本数118287 - 验证集：字节数10358524，样本数5000 下载大小：130923096字节，数据集总大小：255401728字节其余32个配置项结构与上述配置一致，仅模型类型、特征维度、下载大小与数据集总大小存在差异，完整配置项包括：clip_b16_openai、clip_b32_datacomp、clip_b32_laion2b、clip_b32_openai、clip_bigg14_laion2b、clip_g14_laion2b、clip_h14_laion2b、clip_l14_336_openai、clip_l14_datacomp、clip_l14_laion2b、clip_l14_openai、dinov2_b14、dinov2_b14_reg、dinov2_g14、dinov2_g14_reg、dinov2_l14、dinov2_l14_reg、dinov2_s14、dinov2_s14_reg、mae_b16、mae_h14、mae_l16、siglip2_b16_256、siglip2_b16_512、siglip2_l16_384、siglip_b16_384、siglip_b16_512、siglip_l16_256、siglip_l16_384、siglip_so400m_384、vit_b16_21k、vit_l16_21k、vit_s16_21k 各配置项的数据文件路径均遵循`{配置名称}/train-*`与`{配置名称}/val-*`的格式。 ### 补充说明本数据集包含针对COCO（Common Objects in Context）2017的批量预处理数据集。基于此，将开展首批宽域视觉Transformer（Vision Transformer，简称ViT）模型的测试工作，涉及模型包括ViT-Beatrix、ViT-Zana、ViT-Beatrix-DualStream、Clip-Vit-Beatrix、GeoVit-Beans等。这类宽域ViT将采用全新的公式形式，旨在融合同时存在于多种不同流形上的结构行为特性。本次实验将采用基于当前最优性能（State-of-the-art，简称SOTA）的成熟流程，该流程源自多种迁移学习理念，并针对地理分形行为进行了适配与调制。此外，本数据集的特征可供有需求的研究者使用。

应用场景：