Into the Technicalities of Photography: an EXIF Data Exploration
The goal of my project is to explore the technical aesthetic preferences of photographers through visualizing multidimensional image metadata. After theoretically examining and pursing photography professionally i realized that with time every photographer adopts a certain aesthetic preference. This image aesthetic depends on the visual aesthetic (composition ,color etc) as well technical aesthetic (depth of field,perspective,blur). Although there has been significant research conducted in understanding visual aesthetics, very little research has been done to understand technical aesthetics.
33 local photographers took part in a user study where they picked out images that matched their aesthetic preference from a flickr image dataset. Multivariate data visualization tools were created (using D3) for photographers to explore image space and gain insight into the technical preferences of the images they selected.
In the recent years, there has been significant interest in the scientific research community to analyze aesthetics using image processing algorithms. However there has been no scientific research in analyzing metadata to understand photographic aesthetics, especially from a photographer’s point of view. The goal of this project is to understand a photographer’s aesthetic preferences by analyzing and visualizing image metadata.
Theoretically, I am interested in understanding how photographers technically analyze a scene and use this knowledge to create consistent work. I believe that a single image doesn’t represent a photographer, but several images can reveal their aesthetics preference. The objective of this project is achieved through the means of a user study where images selected by the photographer are spatially reorganized using the EXIF data. Using this method, the EXIF data can be visualized in a meaningful manner to understand the photographer’s image aesthetics.
Note: The interfaces are currently optimized for 1920*1080 displays for my thesis presentation, i will be updating them soon!
Interface #1 Kohonen Self Organizing Maps
Artificial Neural Network for visualizing multidimensional data.The images selected by the user are spatially organized based on the EXIF parameters.
- Displaying similarities
- Reduces dimensions
- Unsupervised Learning
Interface #2 Parallel Coordinates
Interface for interacting with multidimensional user study data.
Every line across the major parallel coordinates describes a user response during the user study. Every line across the smaller parallel coordinate describes the image EXIF data.
This work is a result of theoretical examination and EXIF data exploration of digital images. The research presents a new approach to image study, which is not content based but instead uses the camera’s technical parameters (embedded in the image) as a source.
Modern digital photography has evolved at an extremely fast pace in the recent years. High performing sensors, coupled with quality optical lenses have allowed photographers to capture and share increasingly detailed images. At the same time, these high-resolution cameras make photography very unforgiving for someone unaware of the technicalities involved. This attention to technical detail comes with practice and evolves over time. I believe that skilled photographers eventually adopt a certain image aesthetic that defines their images. Image aesthetics are a combination of visual aesthetics and technical aesthetics. Hence “aesthetic” isn't just based on color, context, content and composition but also includes “technical gestalt” which is a result of the camera and lens parameters (i.e. Aperture, ISO, Shutter Speed and Focal Length). From my experience as a professional photographer I have learnt that brilliant images are not just a result of content composition but also technical execution.
The goal of my research is to understand the photographer and their technical aesthetic preferences. The underlying assumption is that every photographer eventually develops a technical gestalt resulting in consistent work that is irrespective of the subject or composition. In other words we tend to develop certain habits for controlling the camera settings, which has a profound effect on the final image. These settings are stored as image metadata (also known as EXIF). For example, for the same scene illumination, photographers may setup the camera differently resulting in very different images. Conversely, if a photographer selects a certain subset of images from an image set based on their preferred aesthetic preferences; I believe that there will be certain degree of correlation between the EXIF data. Understanding and visualizing this relationship will help photographers make informed decisions while creating consistent work.
To collect relevant data, I conducted a user study with local Santa Barbara photographers and correlated my survey findings with the data gathered by an image selection application. The user study consists of a pre user study, image selection, and a post study questionnaire. The data from the user study was combined to create online interactive visualizations, which visually demonstrate every photographer’s unique perspective, image preferences and its correlation to EXIF information.
Image Selection Application
Self Organizing Maps Interaface
Parallel Coordinates Interface
Highcharts for User Study
Keywords: Data Visualization, User Research, Interaction Design, Web development