acervos-digitais/ft-0915
收藏Hugging Face2024-10-14 更新2025-11-03 收录
下载链接:
https://hf-mirror.com/datasets/acervos-digitais/ft-0915
下载链接
链接失效反馈官方服务:
资源简介:
---
size_categories:
- n<1K
task_categories:
- object-detection
pretty_name: Awning, Balcony, Ramp, Sign
dataset_info:
features:
- name: image_id
dtype: int64
- name: image
dtype: image
- name: image_filename
dtype: string
- name: width
dtype: int64
- name: height
dtype: int64
- name: objects
sequence:
- name: bbox_id
dtype: int64
- name: category
dtype:
class_label:
names:
'0': awning
'1': balcony
'2': ramp
- name: bbox
sequence: int64
length: 4
- name: super_category
dtype:
class_label:
names:
'0': object
- name: area
dtype: int64
- name: is_crowd
dtype: bool
splits:
- name: train
num_bytes: 27908679.0
num_examples: 298
- name: test
num_bytes: 7607168.0
num_examples: 74
download_size: 35308334
dataset_size: 35515847.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
---
Mini dataset for architectural object detection fine-tuning.
Manually selected from results of a zero-shot object detection model.
There are 3 positive classes, with 100-150 examples of each.
Objects of interest:
- awnings
- balconies
- pedestrian ramps
size_categories:
- 样本规模<1000
task_categories:
- 目标检测(object-detection)
pretty_name: 遮阳篷、阳台、坡道、标识牌
dataset_info:
features:
- name: 图像ID(image_id)
dtype: 64位整数
- name: 图像(image)
dtype: 图像类型
- name: 图像文件名(image_filename)
dtype: 字符串
- name: 宽度(width)
dtype: 64位整数
- name: 高度(height)
dtype: 64位整数
- name: 目标集(objects)
sequence:
- name: 边界框ID(bbox_id)
dtype: 64位整数
- name: 类别(category)
dtype:
class_label:
names:
'0': 遮阳篷(awning)
'1': 阳台(balcony)
'2': 坡道(ramp)
- name: 边界框(bounding box,简称bbox)
sequence: 64位整数序列
length: 4
- name: 超类别(super_category)
dtype:
class_label:
names:
'0': 物体(object)
- name: 面积(area)
dtype: 64位整数
- name: 是否为密集目标(is_crowd)
dtype: 布尔类型
splits:
- name: 训练集(train)
num_bytes: 27908679.0
num_examples: 298
- name: 测试集(test)
num_bytes: 7607168.0
num_examples: 74
download_size: 35308334
dataset_size: 35515847.0
configs:
- config_name: 默认配置(default)
data_files:
- split: 训练集(train)
path: data/train-*
- split: 测试集(test)
path: data/test-*
本微型数据集为面向建筑结构目标检测微调的专用数据集。
其样本由零样本目标检测模型(zero-shot object detection model)的推理结果手动筛选得到。
数据集共包含3个正样本类别,每个类别包含100至150个标注样本。
需关注的目标类别如下:
- 遮阳篷(awning)
- 阳台(balcony)
- 人行坡道(pedestrian ramp)
提供机构:
acervos-digitais



