Zachary Pezzementi and Trenton Tabor from the National Robotics Engineering Center (NREC) at Carnegie Mellon University in Pittsburgh, USA presented our first re-booted webinar on Friday, November 2, 2018, 10am EDT (GMT -4:00).
Slides are available: pptx [215 MB] or pdf [165 MB]. *Note: embedded videos in the .pdf are not playing Linux (evince, okular), but are in Windows. FYI.
The recording of the webinar is here, .mp4, [187MB] and also on YouTube.
Talk details
Title: Comparing apples and oranges: Off‐road pedestrian detection on the National Robotics Engineering Center agricultural person‐detection dataset
Abstract: Person detection from vehicles has made rapid progress recently with the advent of multiple high‐quality datasets of urban and highway driving, yet no large‐scale benchmark is available for the same problem in off‐road or agricultural environments. Here we present the publicly available National Robotics Engineering Center (NREC) Agricultural Person‐Detection Dataset to spur research in these
An example image from the NREC Person Detection Dataset, where a person is squatting in an orchard row.
environments. It consists of labeled stereo video of people in orange and apple orchards taken from two perception platforms (a tractor and a pickup truck), along with vehicle position data from RTK GPS. We define a benchmark on part of the dataset that combines a total of 76k labeled person images and 19k sampled person‐free images. The dataset highlights several key challenges of the domain, including varying environment, substantial occlusion by vegetation, people in motion and in nonstandard poses, and people seen from a variety of distances; metadata are included to allow targeted evaluation of each of these effects. We present baseline detection performance results for some leading approaches from urban pedestrian detection along with our own convolutional neural network approach that benefits from the incorporation of additional image context. We show that the success of existing approaches on urban data does not transfer directly to this domain. Finally, we briefly review some work that makes use of the dataset for evaluating the robustness of perception systems to challenging conditions.
Supplemental materials
Dataset links:
https://data.nal.usda.gov/dataset/national-robotics-engineering-center-agricultural-person-detection-dataset
https://www.nrec.ri.cmu.edu/solutions/agriculture/other-agriculture-projects/human-detection-and-tracking.html
Dataset paper:
Another example image from the NREC Person Detection Dataset, where a person is squatting in trees to the left of a vehicle.
Zachary Pezzementi, Trenton Tabor, Peiyun Hu, Jonathan K. Chang, Deva Ramanan, Carl Wellington, Benzun P. Wisely Babu, and Herman Herman. "Comparing Apples and Oranges: Off-Road Pedestrian Detection on the NREC Agricultural Person-Detection Dataset". Journal of Field Robotics (JFR), 35:4, June 2018. pp.545-563.
Publisher: https://onlinelibrary.wiley.com/doi/abs/10.1002/rob.21760
arXiv: https://arxiv.org/abs/1707.07169
Robustness testing work using dataset (an additional paper):
Zachary Pezzementi, Trenton Tabor, Samuel Yim, Jonathan Chang, William Drozd, David Guttendorf, Michael Wagner, and Philip Koopman. "Putting Image Manipulations in Context: Robustness Testing for Safe Perception". In IEEE International Symposium on Safety, Security, and Rescue Robotics, Philadelphia, PA, August 2018.
PDF: http://cs.jhu.edu/~zap/papers/pezzementi18_perception_robustness_testing.pdf
Biographical information
Zachary Pezzementi is a lead robotics engineer at the National Robotics Engineering Center at Carnegie-Mellon University. His research interests focus on robotic perception and machine learning, with applications in robotics.
Zach received his Bachelor's degrees in Computer Science and Engineering from Swarthmore College in 2005. He received his Masters and Ph.D. in Computer Science from Johns Hopkins University, where he worked on haptics and visual tracking for surgical robotics. He defended his thesis on "Object Recognition with Tactile Force Sensing" in 2011 and then joined NREC, where he has worked on projects in mobile robotics, agricultural automation, and robotic safeguarding.
Trenton Tabor is a Senior Robotics Engineer at the National Robotics Engineering Center (NREC). He has been part of NREC for four years and studied robotics for several years before that. He has expertise in applying computer vision and machine learning to robotics problems. His recent work has been on testing person detection algorithms in off-road settings. His interest in robotics extends to real-world applications of sensing and manipulation. He is also a volunteer with robotics education organizations and an advocate for diversity in robotics and technology in general.