PASCALRAW: Raw Image Database for Object Detection

To our knowledge, this work presents the first largescale RAW image database for object detection. It contains 4,259 annotated RAW images, with 3 annotated object classes (car, person, and bicycle), and is modeled after the PASCAL VOC database [1]. All annotations were made in accordance with the original PASCAL VOC guidelines [2] and consist of 1,765 cars, 4,077 persons, and 708 bicycles. Unlike processed JPEG images, RAW images are composed of high-precision (12-bit) raw photosensor outputs, which are proportional to scene illuminance. Therefore, this database can be used to simulate the effect of algorithmic hardware implementations—such as embedded feature extraction—at the image sensor or readout level on end-to-end object detection performance. All images represent daytime scenes in Palo Alto and San Francisco, and were captured using a Nikon D3200 DSLR camera. The original full-resolution RAW and JPEG images are provided along with 10X downscaled 600x400 images with corresponding annotations. Additional database details can be found in Appendix A of [3], and instructions on how to use the database are provided in README.txt.

For researchers who prefer not to download the complete ~150 GB dataset provided here, please download the README.txt file for instructions on how to access a smaller ~1GB directory containing only the 10X downscaled images to start.

REFERENCES
[1] Everingham, M. and Van~Gool, L. and Williams, C. K. I. and Winn, J. and Zisserman, A., "The PASCAL Visual Object Classes Challenge 2012 (VOC2012) Results," [Online] Available: http://www.pascal-network.org/challenges/VOC/voc2012/workshop/index.html
[2] Accessed on Sep. 16, 2016. [Online]. http://host.robots.ox.ac.uk/pascal/VOC/voc2007/guidelines.html
[3] A. Omid-Zohoor; C. Young; D. Ta; B. Murmann, "Towards Always-On Mobile Object Detection: Energy vs. Performance Tradeoffs for Embedded HOG Feature Extraction," in IEEE Transactions on Circuits and Systems for Video Technology, vol. PP, no. 99, pp. 1-14.