Human-scale Localization Platform (HULOP)

HULOP help researchers and developers invent new technologies to create engaging experiences for the visually impaired who lack well-refined navigation applications to move freely, without relying on outside assistance

for Blind User

NavCog App

HULOP project provides an app called "NavCog", which aims at realizing easy-to-setup indoor and outdoor seamless navigation for the blind.

Learn more about NavCog

Note: video recognition part is not included in "NavCog" app.

for Supporter

Build your map with Bluetooth beacons to support blind users on your place such as university campus, hospital, and shopping mall. It is not difficult to start.

Quick Start Guide (English, Japanese)

Public Maps for NavCog

for Developer

"This open platform will help expand the horizon for collaboration around the world to open up the new real world accessibility era for the blind in the near future.”

Fork HULOP on GitHub

for Researcher

Localization with BLE beacon: This is a Bluetooth beacon indoor localization and a PDR algorithm components to locate smartphones. It can estimate probability of location of the device from signal strength of beacons distributed along with the indoor routes.
Localization with vision: This tool is able to build 3D model from image taken by a smartphone and also to locate the smartphone in the 3D model.
Blind navigation UI: The user interface provides auditory access of topological map and accessibility information indoor / outdoor environment. Navigation command has been optimized based on user experiments with blind users.
Activity recognizer: The first set of software in this repository includes activity recognition software to extract hand and object interaction from videos of the first-person point-of-view. [ICIP paper]

Related Projects

PerPix: PerPix is a per-pixel regression algorithm which has been develop for detecting hand regions using a wearable camera. The algorithm examines a small region around each pixel, extracts various features and makes a decision using a random forest regressor. This online release is a simplified version of a previously published [CVPR paper]. Instead of learning scene clusters, it treats each image as its own scene cluster.
Acoustic Location Processing System (ALPS): ALPs is an indoor ultrasonic ranging technique that can be used to localize modern mobile devices like smartphones and tablets. The method uses a communication scheme in the audio bandwidth just above the human hearing frequency range where mobile devices are still sensitive.
ZFace: ZFace is a real-time face alignment software which accomplishes dense 3D registration from 2D videos and images without requiring person-specific training. It estimates the 2D and 3D locations of fiducial landmarks, calculates a dense 3D structure of the face and provides accurate 3D head-pose estimates. [FG paper]