Places: An Image Database for Deep Scene Understanding

Abstract

The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach nearhuman semantic classification at tasks such as object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories and attributes, comprising a quasi-exhaustive list of the types of environments encountered in the world. Using state of the art Convolutional Neural Networks, we provide impressive baseline performances at scene classification. With its high-coverage and high-diversity of exemplars, the Places Database offers an ecosystem to guide future progress on currently intractable visual recognition problems.

Comparison of Visual Datasets for Machine Learning

“…While classification tasks require a single ground-truth label, using images with many labels gives the data more context. Furthermore, while large classification datasets exist, such as Places2 [25] with more than 10 million images and Tiny Images [26] containing 80 million image, the CAM2 system can retrieve more that 95 million images in a single day. Moreover, the data from network cameras can provide longterm observations.…”

Squeezeposenet: Image Based Pose Regression With Small Convolutional Neural Networks for Real Time Uas Navigation

“…For pose estimation we added two dense layers ('x' and 'q') with 3 respectively 4 neurons. All weighting parameters up to the 'pool10'-layer are initialized from a pre-trained version of SqueezeNet that was trained on a subset of the Places365 data set (Zhou et al, 2016). The additional layers are pre-trained using the Shop Façade data set and later on our test set of the Atrium.…”

“…It was shown that for pose regression, training on the Places data set (Zhou et al, 2014(Zhou et al, , 2016 leads to further improvement in terms of accuracy (Kendall et al, 2015), as it is a more suitable data set for pose regression. Since that, training was subsequent carried out on the Places data set.…”

DTCTH: a discriminative local pattern descriptor for image classification

“…Eventually, a model is built that can later be used to solve a particular problem which is known as fine tuning. Among the existing popular models, AlexNet [70], Places-CNN [71], and VGG_S [72] are widely used because they cover diversified applications. Despite the gain of popularity of deep learning, it is very computation intensive and requires expensive hardware and large set of training data.…”