Image aesthetics classification using deep features and image category
Document Type
Conference Proceeding
Publication Title
2021 36th International Conference on Image and Vision Computing New Zealand (IVCNZ)
Volume
2021-December
Publisher
IEEE
School
School of Science / Graduate Research
RAS ID
40627
Abstract
Selection of aesthetic images is important in any application where a visual presentation is required, including industries such as e-commerce, tourism, and real-estate. With large datasets of pictures typically available, manual selection of an appropriate image is a difficult task. Automated aesthetic classification aims to address this problem. In our approach, we constrain the problem to determine aesthetics of an image based on knowing what category of image it is, as different image categories have different associated aesthetics. For example, different considerations for camera placement, lighting, etc. are made when capturing a human portrait, compared to a photo of a landscape to produce an aesthetic result. Our approach integrates an individual classifier for each image category type. Image features used by the classifier are extracted from three off-the-shelf pre-trained convolutional deep neural networks. Our approach achieves a higher classification accuracy compared to classifiers that do not take category into account. Further, our off-the-shelf deep feature-based classifiers show improved accuracy over results reported in the literature.
DOI
10.1109/IVCNZ54163.2021.9653375
Access Rights
subscription content
Comments
Maqbool, H., & Masek, M. (2021, December). Image aesthetics classification using deep features and image category [Paper presentation]. 2021 36th International Conference on Image and Vision Computing New Zealand (IVCNZ), Tauranga, New Zealand.
https://doi.org/10.1109/IVCNZ54163.2021.9653375