Computer Laboratory

Projects

Teaser figure
Distribution of pixel values under different adversarial illumination conditions for all scenes in the dataset. The HDR condition resulted in the widest histograms, often with two distinct modes. The histograms under the night illumination are shifted to the left, making images darker and more affected by noise. Finally, the histograms are shifted to the right in the glare condition due to the scattered light.

Abstract

Benchmark datasets used for testing computer vision methods often contain little variation in illumination. The methods that perform well on these datasets have been observed to fail under challenging illumination conditions encountered in the real world, in particular when the dynamic range of a scene is high. We present a new dataset for evaluating computer vision methods in challenging illumination conditions such as low-light, high dynamic range, and glare. The main feature of the dataset is that each scene has been captured in all the adversarial illuminations. Moreover, each scene includes an additional reference condition with uniform illumination, which can be used to automatically generate labels for the tested computer vision methods. We demonstrate the usefulness of the dataset in a preliminary study, by evaluating the performance of popular face detection, optical flow, and object detection methods under adversarial illumination conditions. We further assess whether the performance of these applications can be improved if a different transfer function is used.

Video

Testing framework

Pipeline and TFs

The test conditions under different illuminations are provided as linear 1936 x 1096 PNG images. We use a simplified ISP pipeline (depicted in the left figure) to transform linear images with a tone-curve or transfer function before supplying them as inputs to different computer vision methods. We also provide python classes to encode linear images using commonly used transfer functions such as the ones present in the plot on the right.

Downloads

Bibtex

    @inproceedings{hanji2021hdr4cv,
      author    = {Hanji, Param and Alam, Muhammad Z. and Giuliani, Nicola and Chen, Hu and Mantiuk, Rafa{\l} K.},
      title     = {HDR4CV: High dynamic range dataset with adversarial illumination for testing computer vision methods},
      journal   = {Journal of Imaging Science and Technology},
      year      = {2021},
      publisher = {Society for Imaging Science and Technology},
      url       = {http://www.cl.cam.ac.uk/research/rainbow/projects/hdr4cv-dataset/},
    }