leonvanbokhorst/fire-havoc-philips-lac-eindhoven

---
license: creativeml-openrail-m
tags:
- fire
- havoc
- eindhoven
- stable diffusion
- fine-tuning
pretty_name: Havoc after the Fire at Philips LAC Eindhoven
size_categories:
- 1K<n<10K
task_categories:
- unconditional-image-generation
language:
- en
---
# Image Dataset Havoc after the Fire at Philips LAC Eindhoven

## Dataset Description

A large fire in the center of Eindhoven, May 14th, 2023. The old Philips Lighting Application Centre was engulfed in flames, resulting in massive smoke clouds. Over a hundred firefighters were deployed, and there was significant disruption in the city center. This is a dataset containing images of the remains of the building two months later. The footage was taken on July 19, 2023.


## Dataset Structure

The dataset consists of 1167 images depicting the aftermath of the fire havoc. It is primarily designed for fine-tuning or training a Stable Diffusion model, although it can be used for other purposes as well. Each original image is divided into five cropped versions with between 2 to 8 additional random detail crops. Approximately 30 percent of the images are flipped horizontally. All images in the dataset have been resized to either 1024 x 1024, 768 x 1024, or 1024 x 768 resolution.

| Description | Value |
|---------------------------------------------------------|----------------------|
| Number of Images | 1167 |
| Purpose | Fine-tuning / Training Stable Diffusion model |
| Image Processing | Original image five-cropped (all corners and center) with added 1-8 random detail crops per original |
| Flipped Images | Approximately 30% |
| Resolutions | Hand picked 1024x1024, 768x1024, 1024x768 |

原始信息汇总

数据集概述：Havoc after the Fire at Philips LAC Eindhoven

数据集描述

• 事件背景：2023年5月14日，Eindhoven市中心发生大规模火灾，旧的Philips Lighting Application Centre被火焰吞噬，产生大量烟雾。超过一百名消防员参与救援，市中心受到严重影响。
• 数据内容：本数据集包含火灾发生两个月后，即2023年7月19日拍摄的建筑物残骸图像。

数据集结构

• 图像数量：1167张
• 主要用途：用于稳定扩散模型的微调或训练
• 图像处理：每张原始图像被分为五个裁剪版本（四个角和中心），并额外添加1至8个随机细节裁剪。约30%的图像进行了水平翻转。
• 分辨率：所有图像均调整为1024x1024、768x1024或1024x768。

数据集详情

描述	值
图像数量	1167
用途	微调/训练稳定扩散模型
图像处理	原始图像五裁剪（所有角落和中心）+ 1-8个随机细节裁剪
翻转图像	约30%
分辨率	1024x1024, 768x1024, 1024x768