Silviase/CapMetaEval
收藏Hugging Face2024-06-10 更新2024-06-12 收录
下载链接:
https://hf-mirror.com/datasets/Silviase/CapMetaEval
下载链接
链接失效反馈官方服务:
资源简介:
---
dataset_info:
- config_name: composite
features:
- name: image_path
dtype: string
- name: scores
struct:
- name: MAGAer13/mplug-owl2-llama2-7b
struct:
- name: visce_2_new
dtype: string
- name: gpt-4o
struct:
- name: visce_2_new
dtype: string
- name: human_correctness
dtype: float64
- name: human_throughness
dtype: float64
- name: liuhaotian/llava-v1.5-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: visce_attr
dtype: string
- name: visce_obj
dtype: string
- name: liuhaotian/llava-v1.5-13b_strong
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-13b_weak
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
splits:
- name: train
num_bytes: 56482995
num_examples: 18641
download_size: 18651996
dataset_size: 56482995
- config_name: flickr8k-expert
features:
- name: image_path
dtype: string
- name: scores
struct:
- name: MAGAer13/mplug-owl2-llama2-7b
struct:
- name: visce_2_new
dtype: string
- name: gpt-4o
struct:
- name: visce_2_new
dtype: string
- name: human
dtype: float64
- name: liuhaotian/llava-v1.5-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: visce_attr
dtype: string
- name: visce_obj
dtype: string
- name: liuhaotian/llava-v1.5-13b_strong
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-13b_weak
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
splits:
- name: train
num_bytes: 39908638
num_examples: 16992
download_size: 11454375
dataset_size: 39908638
- config_name: pascal-50s
features:
- name: image_path
dtype: string
- name: scores
struct:
- name: MAGAer13/mplug-owl2-llama2-7b
struct:
- name: visce_2_new_p
dtype: string
- name: gpt-4o
struct:
- name: visce_2_p
dtype: string
- name: label
dtype: int64
- name: liuhaotian/llava-v1.5-13b
struct:
- name: base_p
dtype: string
- name: captions_p
dtype: string
- name: references_p
dtype: string
- name: visce_2_new_p
dtype: string
- name: visce_2_p
dtype: string
- name: visce_attr_p
dtype: string
- name: visce_obj_p
dtype: string
- name: liuhaotian/llava-v1.5-13b_strong
struct:
- name: visce_2_new_p
dtype: string
- name: liuhaotian/llava-v1.5-13b_weak
struct:
- name: visce_2_new_p
dtype: string
- name: liuhaotian/llava-v1.5-7b
struct:
- name: base_p
dtype: string
- name: captions_p
dtype: string
- name: references_p
dtype: string
- name: visce_2_new_p
dtype: string
- name: visce_2_p
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-13b
struct:
- name: base_p
dtype: string
- name: captions_p
dtype: string
- name: references_p
dtype: string
- name: visce_2_new_p
dtype: string
- name: visce_2_p
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-7b
struct:
- name: base_p
dtype: string
- name: captions_p
dtype: string
- name: references_p
dtype: string
- name: visce_2_new_p
dtype: string
- name: visce_2_p
dtype: string
- name: category
dtype: int64
splits:
- name: train
num_bytes: 2496675
num_examples: 4000
download_size: 801506
dataset_size: 2496675
- config_name: thumb
features:
- name: image_path
dtype: string
- name: scores
struct:
- name: MAGAer13/mplug-owl2-llama2-7b
struct:
- name: visce_2_new
dtype: string
- name: gpt-4o
struct:
- name: visce_2_new
dtype: string
- name: gpt4o
struct:
- name: base
dtype: string
- name: visce_2
dtype: string
- name: human_p
dtype: float64
- name: human_r
dtype: float64
- name: human_total
dtype: float64
- name: liuhaotian/llava-v1.5-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: visce_attr
dtype: string
- name: visce_obj
dtype: string
- name: liuhaotian/llava-v1.5-13b_strong
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-13b_weak
struct:
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.5-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-13b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
- name: liuhaotian/llava-v1.6-vicuna-7b
struct:
- name: base
dtype: string
- name: captions
dtype: string
- name: references
dtype: string
- name: visce_2
dtype: string
- name: visce_2_new
dtype: string
splits:
- name: train
num_bytes: 9054835
num_examples: 2500
download_size: 3090677
dataset_size: 9054835
configs:
- config_name: composite
data_files:
- split: train
path: composite/train-*
- config_name: flickr8k-expert
data_files:
- split: train
path: flickr8k-expert/train-*
- config_name: pascal-50s
data_files:
- split: train
path: pascal-50s/train-*
- config_name: thumb
data_files:
- split: train
path: thumb/train-*
---
The dataset contains multiple configurations, each detailing features and scores related to image analysis. It includes image paths and scores from various models and human evaluations. Each configuration has specific data files and sizes for training data.
提供机构:
Silviase
原始信息汇总
数据集概述
数据集配置
-
composite
- 特征:
image_path: 字符串类型scores: 结构体类型,包含以下字段:MAGAer13/mplug-owl2-llama2-7b: 结构体类型,包含visce_2_new字段,字符串类型gpt-4o: 结构体类型,包含visce_2_new字段,字符串类型human_correctness: 浮点数类型human_throughness: 浮点数类型liuhaotian/llava-v1.5-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.5-13b_strong: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-13b_weak: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-7b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-7b: 结构体类型,包含多个字段,均为字符串类型
- 分割:
train: 包含 18641 个样本,占用 56482995 字节
- 下载大小: 18651996 字节
- 数据集大小: 56482995 字节
- 特征:
-
flickr8k-expert
- 特征:
image_path: 字符串类型scores: 结构体类型,包含以下字段:MAGAer13/mplug-owl2-llama2-7b: 结构体类型,包含visce_2_new字段,字符串类型gpt-4o: 结构体类型,包含visce_2_new字段,字符串类型human: 浮点数类型liuhaotian/llava-v1.5-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.5-13b_strong: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-13b_weak: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-7b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-7b: 结构体类型,包含多个字段,均为字符串类型
- 分割:
train: 包含 16992 个样本,占用 39908638 字节
- 下载大小: 11454375 字节
- 数据集大小: 39908638 字节
- 特征:
-
pascal-50s
- 特征:
image_path: 字符串类型scores: 结构体类型,包含以下字段:MAGAer13/mplug-owl2-llama2-7b: 结构体类型,包含visce_2_new_p字段,字符串类型gpt-4o: 结构体类型,包含visce_2_p字段,字符串类型label: 整数类型liuhaotian/llava-v1.5-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.5-13b_strong: 结构体类型,包含visce_2_new_p字段,字符串类型liuhaotian/llava-v1.5-13b_weak: 结构体类型,包含visce_2_new_p字段,字符串类型liuhaotian/llava-v1.5-7b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-7b: 结构体类型,包含多个字段,均为字符串类型
category: 整数类型
- 分割:
train: 包含 4000 个样本,占用 2496675 字节
- 下载大小: 801506 字节
- 数据集大小: 2496675 字节
- 特征:
-
thumb
- 特征:
image_path: 字符串类型scores: 结构体类型,包含以下字段:MAGAer13/mplug-owl2-llama2-7b: 结构体类型,包含visce_2_new字段,字符串类型gpt-4o: 结构体类型,包含visce_2_new字段,字符串类型gpt4o: 结构体类型,包含多个字段,均为字符串类型human_p: 浮点数类型human_r: 浮点数类型human_total: 浮点数类型liuhaotian/llava-v1.5-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.5-13b_strong: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-13b_weak: 结构体类型,包含visce_2_new字段,字符串类型liuhaotian/llava-v1.5-7b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-13b: 结构体类型,包含多个字段,均为字符串类型liuhaotian/llava-v1.6-vicuna-7b: 结构体类型,包含多个字段,均为字符串类型
- 分割:
train: 包含 2500 个样本,占用 9054835 字节
- 下载大小: 3090677 字节
- 数据集大小: 9054835 字节
- 特征:



