Skip to main content

EDGE: The 'Eye in the Woods' Ioannis Kakadiaris, PhD | University of Houston


Goal: To create a system to automatically detect and match human subjects from trail camera images. 

Abstract: Trail cameras can provide images of people crossing locations where cameras have been placed. EDGE aims to analyze trail camera images to detect the presence of people, detect and recognize faces that can be matched to known traffickers, count the number of people, and detect carry-load (gun, bag) to characterize individuals. The innovativeness of the work lies in the fact that it will utilize existing source infrastructure, (i.e., the network of trail cameras), and analyze both daylight visible (VIS) and nighttime near-infrared (NIR) images.

The originality of the proposal is that it will create (i) an integrated approach to image-based analysis for human trafficking detection and (ii) a system that will provide meta-information to help the operators understand human trafficking on the larger scale, reveal interdependencies between actors and places and derive possible patterns of movement and detailed activity parameters, such as carry load. The proposed system will be designed as a scalable cloud-driven prototype to facilitate quick and straightforward uptake for the transition to industry-level settings. It is anticipated that the research output of the EDGE project will result in the successful detection and identification of person-of-interest and provide data-driven accurate insight into the whereabouts and activity patterns of human traffickers.

Furthermore, the EDGE project targets scientific advances in (1) pedestrian detection from VIS and NIR images, (2) matching a set of images to a single gallery image, and (3) cross-domain facial matching for VIS and NIR images.

2019 Annual Meeting Presentation

Publication | SSFD+: A Robust Two-Stage Face Detector (Lei Shi, Xiang Xu, Ioannis A. Kakadiaris)

Publication | On the Importance of Feature Aggregation for Face Reconstruction (X. Xu, H. Le, I.A. Kakadiaris)

Publication | Illumination-invariant Face Recognition with Deep Relit Face Images (H. Le, I.A. Kakadiaris)

Publication | Open Source Face Recognition Performance Evaluation Package (X. Xu, H. Le, I.A. Kakadiaris)