Search engine giant Google has released a data set with 5 million images and 200,000 landmarks for training AI models, which can be used by the machine learning and deep teaching developers to train their AI program. The Google AI research division’s main work was to develop a model to accurately recognize the landmark in the image with great accuracy. For example their goal was to distinguishing Niagara Falls from just any waterfall and so on.
Google AI research team released Google-Landmarks, landmarks data set it claimed at the time was the world’s largest, hosted two competitions namely Landmark Recognition 2018 and Landmark Retrieval 2018. These two competitions was very popular and more than 500 machine learning experts participated in this competition.
This competition was very popular and more than 500 machine learning experts used the Google-Landmarks to develop and train the model. The accuracy of most of the model developed and trained using this data set was good.
Now, today in a landmark decision Google open-sourced their famous Google-Landmarks-v2 dataset that comes with 5 million images and 200,000 landmarks. Developers around the world can use this dataset to train their model for recognizing the objects through their application.
Now two new challenges, the Landmark Recognition 2019 and Landmark Retrieval 2019 are launched on the Kaggle, where developers can participate. Google also released the source code and model for Detect-to-Retrieve, a framework for regional image retrieval.
With the release of these image datasets developers can more powerful applications for AI and computer vision. A well trained model can detect the place correctly from the given image dataset. This type of application can be used to find out the place detail from a given photograph. The well trained model can pick-up the background image from a given photograph to recognize the place.
According to Google AI software engineers Bingyi Cao and Tobias Weyand, Image recognition and image retrieval methods request a large-dataset to train the model. Here it is require to train the system with variety of landmarks in order to train better and more robust systems. He further added that “We hope that this dataset will help advance the state-of-the-art in instance recognition and image retrieval.”
According to Bingyi and Weyand, the dataset of Google-Landmarks-v2 contains very large number of photos and it counts over 5 million images; where more than 200,000 are of different landmarks collected from photographers around the world.
The open sourcing of this data set is another milestone towards development of artificial intelligence system which can detect the place from a given image. This dataset contains images collected from many photographers around the world.
