Watching the watchers: Camera identification and characterization using retro-reflections - Dataset
Data files
Mar 19, 2024 version files 12.45 GB
-
README.md
1.64 KB
-
Watching_the_Watchers_data.zip
12.45 GB
Abstract
A focused imaging system such as a camera will reflect light directly back at a light source in a retro-reflection (RR) or cat-eye reflection. RRs provide a signal that is largely independent of distance providing a way to probe cameras at very long ranges. We find that RRs provide a rich source of information on a target camera that can be used for a variety of remote sensing tasks to characterize a target camera including predictions of rotation and camera focusing depth as well as cell phone model classification. We capture three RR datasets to explore these problems with both large commercial lenses and a variety of cell phones. This repository contains time-synced videos from the perspective of both a retro-reflective probe and a target camera that can be used to train algorithms for different remote sensing tasks. We include a dataset for cellphone classification, target camera rotation prediction, and target camera focusing depth prediction.
README: Watching the watchers: Camera identification and characterization using retro-reflections - Dataset
https://doi.org/10.5061/dryad.6t1g1jx64
Description of the data and file structure
This dataset was used to generate results in the paper titled "Watching the watchers: Camera identification and characterization using retro-reflections"
The data folder contains the datasets for the Phone classification and Focus_and_rotation experiments.
data\Focus_Rotation\raw_data\angles: contains the paired videos from the viewpoint of the probe and target camera. The target camera video is used to calculate the rotation angle of the target camera. The probe video is the view from the probe and can be used to predict the target camera's rotation.
data\Focus_Rotation\raw_data\focus contains videos for different settings of a commercial lens on a rotating target camera.
data\PhoneData contains the data relevant for the phone classification experiment. Trimmed_Phone_Videos and Trimmed_Probe_Videos contain the paired videos from the target phone and the probe. Phone_data\imgs contains the background image that is subtracted from the probe videos to remove static background. data\PhoneData\data is left empty and is used to save intermediate data in the linked code repository.
Focus (object) distances are in meters and focal lengths are in millimeters.
Code
Code in this project is written in python and contains some jupyter notebook (ipynb files). There are multiple README files throughout the project code to explain which files replicate different results in the paper.