- 1 Introduction
- 2 Computer Vision Datasets
- 3 Conclusion
With the advent of artificial intelligence and machine learning, there has been an attempt to develop those unique skills in machines that human beings possess in the form of a gift. One of those unique skills is the power of vision. To achieve this, one of the branches of artificial intelligence i.e. computer vision is providing the impetus for this purpose. In this article, we will cover some of the most useful computer vision datasets that you can use for your projects.
Computer Vision Datasets
MNIST database contains handwritten digits with a large set of training and testing examples. The dataset is already normalized i.e. cleaned and it has been optimized for producing the best possible results. It is a very famous dataset that is generally considered useful for beginners because it helps in building a very basic computer vision project of image recognition.
2Dogs & Cats Images Dataset
A very famous and useful dataset for data science aspirants hosted over Kaggle. This dataset contains images of dogs and cats. The dataset is divided into training and testing subsets, this bifurcation of dataset helps in training and testing of the machine learning models. With the help of this dataset, we can also differentiate dog and cat breeds using classification techniques.
A character recognition database with symbols of two different languages i.e. English and Kannada. This dataset has over 74000(74K) images in it and that’s self-explanatory for the name of the dataset. Within this dataset, there are other sub-sections which divides the characters. Based on the requirement of the project, relevant datasets can be downloaded.
A huge image database created by Stanford University where a network of images creates a complete database. Here in this database, there are nodes that consist of more than hundreds and thousands of images.
The dataset can be downloaded in various ways, you can download image URLs, original images and object bounding boxes can be downloaded. There is also an ImageNet API that can expedite this process of downloading.
This is a very good source of different kinds of datasets. Here at this dataset junction, you’ll be provided with the suggestions for projects which can be built using a particular dataset. The salient feature of this source is the documentation of each dataset. Each dataset source is extremely large and can help you to learn a lot by building projects. There are some datasets that consist of ready-made computer vision models as well.
COCO or Common Objects in Context dataset contains a large collection of images. These images are divided into various categories like object categories and stuff categories. The COCO dataset is sponsored by world-renowned organizations such as Facebook and Microsoft.
7CAVE Databases by Columbia University
There is a wide array of databases hosted and provided by Columbia University. The main feature of this source is the variety of datasets that are available for use, you can find an eye-gaze dataset, public figures face database and weather images database. Another useful feature is the documentation of each dataset with information about image count statistics and collection procedure
Visual Genome dataset is an initiative to connect two different sub-fields of Artificial Intelligence, i.e. Computer Vision and Natural Language Processing. The aim of the visual genome dataset is to connect structured image concepts to natural language. This is a humongous dataset with over a million images, over 5 million visual question answers and other pertinent resources. The dataset has various options to download along with this, there is a tutorial for learning purposes.
This dataset is created by a collaboration of the tech giant Google along with Carnegie Mellon University and Cornell University. The dataset comprises almost 9 million images. These images are annotated over 6000 categories, thus making the dataset much more useful.
10Labeled faces dataset
A dataset that comprises different face orientation images of people around the world. It is created and hosted by the University of Massachusetts Amherst. This image collection has almost 13K images, collected from the web. The people whose face images were taken have been mentioned in the dataset, so you can download the images easily as they are organized in an alphabetical manner.
11Indoor Scene Recognition
Generally, we find a dataset of outdoor images but here, the Massachusetts Institute of Technology has compiled an indoor scene images dataset with 67 Indoor categories, and a total of 15620 images. The number of images varies across categories, but there are at least 100 images per category. All images are in jpg format. There is also a subset of images that are annotated and segmented.
12Cancer Images dataset
This is a source of large datasets with images of different types of cancer such as breast cancer, lung cancer, and brain cancer. For each dataset, there is a summary that tells us about the dataset, we have the option of downloading the dataset with different image types such as tissue slide images, DICOM images (a type of image used in the medical industry) and also through API.
So, we have reached the end of this article where we saw plenty of computer vision datasets. I hope you will find these datasets useful for your projects.