# Earthquake Road Crack Detection — Data

来源：https://www.kaggle.com/competitions/earthquake-road-crack-detection/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

On February 6, 2023, a major earthquake occurred in Turkey. One of the biggest problems after the earthquake was the cracks and deterioration in the roads. Therefore, there were many problems in the delivery of aid. 

The "Earthquake Road Crack Detection" competition is a computer vision challenge that aimed at developing algorithms to immediately and automatically detect roads damaged by aerial imagery after affected by earthquakes. The goal of the competition is to build a model that can accurately identify and locate cracks in the images, which can help in the assessment of the damage caused by earthquakes and the planning of repair and maintenance activities.

We are starting this project based on volunteerism to do everything in our power after the earthquake. We will present the project to the authorities and create a text with the names of everyone who participated.T

Not only for the earthquake in Turkey, but we will also share the model that emerged for use in many countries as open source. We are also working on this project ourselves. While doing this, we thought of the idea of why not work together with everyone the rest of the world and that's why we thought of opening this competition.The winning solutions will not only demonstrate the cutting-edge technology in computer vision, but also have the potential to impact real-world applications in the field of road assessment and maintenance.

We already thank everyone who voluntarily participates in this competition. And we share the problems related to our topic.

Step 1: Data
The data is very dirty and we collected it ourselves. We tried to filter it as much as we could, but there is still a lot of work to do. But this can also be considered as a challenge in the competition. Collecting more data/images, filtering and labeling is also a big challenge and we take this as part of the competition. Also, if we can improve this data, it can be turned into a great data set that everyone can use.

There are 2 folders in the data, "filtered" and "unfiltered". In these two folders, there are two folders named "neg"(negative) and "pos"(positive). The "unfiltered" folder contains the unfiltered images that we downloaded from the internet. In the "filtered" folder, there are filtered, resized, jpg formatted images. There are craked road images in the "neg" file, and regular road images with in the "pos" file.

If the images are not enough, you can use google/yahoo/bing/yandex images for data collection. There are browser extension for batch downloading. You can also use different keywords such as earthquake ground crack, road crack while searching. Also, searching in different languages can help you find different images.

Resize: Everyone may want to implement the resize themselves, that why we share images with original sizes. You can do it with photo editing applications has batch option or can with Python.

File format/type has been converted to JPG format (only for the files within the filtered folder).

Size = pos: 2556 images neg: 4000  images  / pos(unfiltered): 0000 , neg(unfiltered): 0000


Step 2: Model
We aim to create a model with as high accuracy as possible. We want to use it to automatically detect cracks in images and videos taken with drones and similar vehicles.

In addition, we have more post-earthquake and projects in the future.

The first one is to detect damaged buildings (cracked).

The second is to detect damaged structures such as bridges, viaducts, dams, etc.

We will collect data as much as possible and open a separate competition for these as well.

If you know any data for these 3 projects, we would be very happy if you shared it with us. We will also share everything with you such as researches, articles etc. Please share with us as well. For example, in wall crack detection ... the data set can be used. When we start that project, we will touch on that subject. Earthquake is a world problem not just Turkey's. Let's make the world a better place together!