Deep learning datasets (traffic signs/flames/handwritten characters/road crack datasets)

Fire and Smoke Image Dataset

This dataset contains early fire and smoke image data. The dataset is early images of fire and smoke captured in real scenes using mobile phones. These images are captured under various lighting conditions (indoor and outdoor scenes), weather, etc. This dataset is well suited for early fire and smoke detection. The dataset can be used for fire and smoke recognition, detection, early fire and smoke detection, anomaly detection, etc. The dataset also includes typical household scenes, such as garbage burning, paper and plastic burning, field crop burning, home cooking, etc.

GTSRB German traffic sign dataset

The German Traffic Sign Benchmark is a multi-class single image classification challenge presented at the 2011 International Joint Conference on Neural Networks (IJCNN). Researchers in relevant fields are invited to participate: the competition is designed to require no special domain knowledge for participants. Our benchmark has the following properties:

1. Single. Image, multi-class classification problems 2.
More than 40 classes
3. Over 50,000 images in total
4. Realistic large database

MNIST Handwritten Digit Image Dataset

The MNIST data set is a handwritten Arabic numeral image recognition data set. The picture resolution is 20x20 grayscale pictures, including '0 - 9' ten groups of handwritten handwritten Arabic numerals. Among them, there are 60000 training samples and 10000 test samples, and the data is the pixel value of the picture. The author has compressed the data set.

CrackForest dataset

The CrackForest dataset is an annotated image database of road cracks, which can roughly reflect the condition of urban pavement