https://www.williamhoff.tech/home daily 1.0 2026-03-11 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606155080085-7B4SCA9J6PTRGWVDO9ZP/Hoff.png Home Researcher, Developer, Educator Hello! I’m William (Bill) Hoff, and I am a researcher, developer, and educator in augmented reality and computer vision. My passion is developing augmented reality and mixed reality technology that empowers people. These technologies superimpose information on top of images of the real world using head-mounted displays or smart phones. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607202970661-IA1OZXWIIPBGPX9MWI48/2019A.jpg Home https://www.williamhoff.tech/projects daily 0.75 2026-03-10 https://www.williamhoff.tech/publications daily 0.75 2026-03-06 https://www.williamhoff.tech/tagging daily 0.75 2026-03-06 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606437996634-0G9JYGDVFK2UEZA3DTIP/motor.png Tagging - I developed this technology while working at DAQRI Corp. in Vienna, Austria. DAQRI wanted object detection capability integrated into their augmented reality “smart glasses”. Specifically, they wanted to be able to “tag” a real-world object with virtual annotations, such as labels and instructions. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606454796000-YBYKOP53MG45M1X9G0MQ/features.png Tagging The method uses a convolutional neural network (CNN) to extract dense features from images. The CNN uses pre-trained weights, so that it doesn’t have to be retrained for each new object. Details are in my patent (see https://patents.google.com/patent/US20200074672A1/en). https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606502497005-RWWUOZIP6ONOYMJW54LI/clusters.png Tagging The system gradually learns a 3D model of the object, even though there is no explicit labeling of which points belong to the object and which points belong to the background. As more and more people use the system, the object appears in different scenes with different backgrounds. Thus, the system learns to ignore points from the background and which points are part of the object. As a result, the recognition model improves over time. https://www.williamhoff.tech/task-guidance daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606512221539-2SD8QV8ZW0KDQCUGZW8C/wireguide.png Task Guidance Overlays show which posts to connect with a wire. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606509765207-X3MD7LTVN3VBNJYGXRRN/atg.png Task Guidance https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606512495191-TN6EWGHUXASOR81WC0EY/copy_machine.png Task Guidance https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606512518676-6SV2DNHJUBALAN6HB8X1/pomdp.png Task Guidance The task of unjamming a copier is modeled as a state machine with actions, observations, and rewards. https://www.williamhoff.tech/feature daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606519135099-50G4ECL23ISXRNY5S97I/VSLAM_patches.jpg Feature Tracking https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606520613680-9I3KZERIGM4TDDM1K778/VSLAMtrajectories.png Feature Tracking https://www.williamhoff.tech/landmark-recognition daily 0.75 2020-12-06 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606583059662-JWPVFP29XJLYHXM1USLZ/vehicle1.jpg Landmark Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607277167799-QJ2BYQ2LFRM1T3BFDS9D/cameras.png Landmark Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606534090945-KDDJM1PNMXQZWKC7AXVI/roxborough.png Landmark Recognition Map showing the path of the vehicle. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606584269722-MVKP7XDX5TKNUL2HPXUC/image-asset.png Landmark Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606585083649-8XCBTVD2OSY3EAC1MZUR/eigenimages.png Landmark Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606586032654-QRH0NL8FVSLYRLU6V8H8/lateral_error.png Landmark Recognition Lateral positioning error vs time, for the landmark recognition method (left) and odometry-only method (right). https://www.williamhoff.tech/chess-recognition daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606587769734-6ABJMCMMLB0GI12JKU3H/chess_calibration.png Chess Recognition Chess boards are widely used for camera calibration, but this is not what we are addressing in this project. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606587803755-Y8B209X8UBR787SDB6OC/chess1.jpg Chess Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606588428650-5JCOW5DWASILS8K922SY/chessboard_finding.png Chess Recognition https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606589915793-PV9972R8JTEMLWIIRC09/chess_recognition.jpg Chess Recognition https://www.williamhoff.tech/pedestrian-detection daily 0.75 2020-11-28 https://www.williamhoff.tech/registration-3d daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606606359310-AFBCK8XWPBBCY029L0ZZ/acrs.jpg Registration of 3D Data https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606607107377-2RD3D7WLG0V15RFN2G0T/DegenerateCase.png Registration of 3D Data A planar surface is degenerate, in that the pose of the sensor is not fully constrained. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606607126395-BLWKW9I4L4KAEEKXEDTR/QuasiDegenerateCase.png Registration of 3D Data Small structures disambiguate the pose, but may be missed in a random sample. We call this “quasi-degenerate”. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606608150169-LTP6QDKJJNIQIDT2ZSRY/table5.4.jpg Registration of 3D Data Probability of algorithm to compute a solution that reduces the error by 80% and 95%. The values are given as percentages. https://www.williamhoff.tech/activity-recognition daily 0.75 2026-03-12 https://www.williamhoff.tech/pose-accuracy daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606612041416-AX3EYVIFUTD5ZN3QTMAG/head_pose1.jpg Pose Accuracy https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1606612069982-UNMMANLATY93LIVY74ET/head_pose2.jpg Pose Accuracy https://www.williamhoff.tech/courses-1 daily 0.75 2020-12-07 https://www.williamhoff.tech/personal daily 0.75 2026-03-04 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607301632851-L906XN32BS376JKXTCY2/SchonbrunnPalace.jpg Personal Schönbrunn Palace near Vienna, Austria. We lived in Vienna from June 2017-November 2018. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607301868391-E8733PR1Q4JI5ZSWHP8H/moab.jpg Personal Mountain biking is incredible in Moab, Utah. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607301923179-5ACGTO7U3K24U7N7F4L5/myAchilles.jpg Personal https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607302108125-8A9VQ3SMEL7FQIRM48I9/maui.jpg Personal Summit of Haleakala peak in Maui, Hawaii. There is a famous astronomical observatory at the top. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607625298812-KVM1J8ZE0XWR36ZO9UIE/ken.jpg Personal https://www.williamhoff.tech/intro-to-computer-vision daily 0.75 2026-03-04 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607204972639-6DZGVCBAHCRBTDXVY2LU/replace1.png Intro To Computer Vision Original image. The goal was to replace the circled sign with another image texture. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607205032067-88UYAPIIQMJQ12Z7XIFM/replace2.png Intro To Computer Vision The new image with the sign replaced by a different image texture. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607207030061-M7QKXXR0E27TI15A814Z/training.png Intro To Computer Vision Training image. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607207049219-BFFJY10ZSPAICL5OWM06/query1.png Intro To Computer Vision Query image, successful detection. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607207091193-M1FRD9W7SWEDQV4Z2TEJ/query2.png Intro To Computer Vision Query image, successful detection. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607207792170-MJXKG4OV5SK2L6DQFYF4/gauges.jpg Intro To Computer Vision https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607207991463-85DL8HHNCMB6NAH3E4TZ/pooltable.jpg Intro To Computer Vision https://www.williamhoff.tech/finding-correspondences daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607051350245-TZ59UMGTJVRNXQV5V75U/correspondences.png Finding Correspondences https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607051640107-5YRJIUQOY5LMISIV8X3C/inliers.png Finding Correspondences Our method finds the inlier matches between the template image (top) and the test image (bottom), and classifies them as coming from two different objects. https://www.williamhoff.tech/contact daily 0.75 2026-03-07 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607200033534-O0OOTH8937DHZNM5AT0Z/astronaut.jpg Contact - Get in Touch! If you are developing AR technology and applications, and need an experienced researcher and developer, contact me. I am interested in part time positions, full time positions and/or consulting work. My email address: william.albert.hoff@gmail.com https://www.williamhoff.tech/advanced-computer-vision daily 0.75 2026-03-04 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607211073786-51AZCWFTDU65MFU38KKS/pavillion1.png Advanced Computer Vision First image, with epipolar lines. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607211117268-XV9AJYLP18UDGWGLJ2BE/pavillion2.png Advanced Computer Vision Second image, with epipolar lines. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607211910466-I1M4G5IVP7LSM8FWM34K/sfm1.png Advanced Computer Vision One of the images, with the “ground control points” marked. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607211928389-DZUNJEYRF2T34YJT43C1/sfm2.png Advanced Computer Vision Reconstructed camera poses and 3D point positions. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607212714765-2F1Z7UKV14WLJGTYKIXR/sign1.png Advanced Computer Vision https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607212761524-WEH7OBBORZQ7HGIBEDAU/sign3.png Advanced Computer Vision https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607213751715-8EW6LIDKBBVCVVHPSHYO/artifact1.png Advanced Computer Vision Given the query image in the top left, the method finds the closest matching images in the database. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607213937022-3WGOFEOF6XEK0JPCX6P9/artifact2.jpg Advanced Computer Vision Detected lines are used to find vanishing points. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607213951660-QUCEINX92S7TCZGFW0QT/artifact3.png Advanced Computer Vision Artifacts are placed in the image. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/7935467d-0b5b-480e-8e45-da7046e123b3/Ruiz-figure1.png Advanced Computer Vision - Make it stand out Input image (left), rectified image (center), solved puzzle (right). https://www.williamhoff.tech/augmented-reality daily 0.75 2026-03-10 https://www.williamhoff.tech/visual-inertial daily 0.75 2021-02-09 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607296925596-MJB88U3V4BJGPV2CSFY6/helmet1.jpg Visual Inertial https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607297534586-EIEUOWXLFC439CCV00SQ/optotrak.jpg Visual Inertial https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607298164272-A2435W2Q7GM4KIJ85GYA/EKF1.png Visual Inertial https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607298421980-PQZNX1S60PLA7KMUP3WA/EKF3.png Visual Inertial Block diagram of head motion filter. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607298551529-LOVOQBKB2EKX5GQL1NDH/EKF2.png Visual Inertial Block diagram of structure estimation filter. https://www.williamhoff.tech/software-engineering daily 0.75 2020-12-07 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1607301060132-9U5SW4RJ9LFI7OSEPFAH/clue_game.png Software Engineering An implementation of the board game “Clue”. https://www.williamhoff.tech/rfk daily 0.75 2020-12-10 https://www.williamhoff.tech/rapid-authoring daily 0.75 2021-03-27 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1614012553135-W7D5D7HLTAJ98RDQRR39/vhs+repair.jpg Rapid Authoring - Imagine being able to see instructions for how to perform a difficult task, displayed right on top of the object you are working on. For example, it could show you the location of a switch, or how to place a tool. This is extremely helpful to guide you through a task that you have never done before, such as operating or repairing some complex equipment. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1614012587371-8T43F4R32LCVXG08KZXM/screwdriver+tag.jpg Rapid Authoring - Augmented reality (AR), running on a smart phone, can show instructions and annotations that are displayed right on top of equipment. These annotations (“tags”) can be 3D models of tools, animated arrows, or text. The key idea is that they are continuously aligned with the object, so that as the person moves around, the tag appears to stay on top of the object. AR task guidance can enable people to do tasks that they have never done before. Research shows that augmented reality can reduce task time (see Henderson, et al) and the number of errors (see Webel, et al). https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/1613947198951-6ASILU3NWLVYPTVG7W11/covariance.png Rapid Authoring The covariance of a tag’s position is represented by a 3D ellipsoid. https://www.williamhoff.tech/digital-twin daily 0.75 2026-03-06 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/6091aaf4-dde6-48ae-956a-57cae81967bf/U-net-arch.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/460aeb7f-77a2-4683-ae65-584ac3820c98/barge-rendered-images.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/94664f97-f6e3-4585-8d86-eb8206536a78/flange-fitting.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/13cd8624-c987-4a0c-86af-400bb7a7a9ee/semantic-classes-legend.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/50e1a4f9-8f6c-4e74-aefb-c2fc6e8087dd/flange-detections.png https://www.williamhoff.tech/object-detection daily 0.75 2026-03-08 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/73391961-34a0-4c40-a197-44f0398f48cd/obj_recog_reference_1.png Object Detection from Reference Images Reference image, with annotations placed by the first user. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/2c00b59d-6e40-43c7-9091-c67736f51b42/obj_recog_reference_3.png Object Detection from Reference Images Matching features on the reference image. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/dc05c56a-bf5b-48da-b9da-763fc137c978/obj_recog_reference_2.png Object Detection from Reference Images Query image, with annotations automatically placed by the system. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/70c94609-3bfb-4cd7-b1b9-ff637d6460ae/obj_recog_reference_4.png Object Detection from Reference Images Matching features on the query image. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/8b8d2df7-bf72-4741-9f69-f1fe7b5169f9/obj_recog_reference_5.png Object Detection from Reference Images Reference image, with annotations placed by the first user. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/9df26e14-528b-4c83-b4a2-ba2fa0604b43/obj_recog_reference_6.png Object Detection from Reference Images Query image, with annotations automatically placed by the system. https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/b4250bbc-8f99-4589-a5ab-65b818cc1eab/object-detection-legend.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/7ff08979-9aa7-4357-b95f-ffa2a4d84b95/reference_images_valve.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/f6ff5e9a-7883-42fc-a0a4-366485a64d55/reference_images_hydrant.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/f6ff5e9a-7883-42fc-a0a4-366485a64d55/reference_images_hydrant.png https://www.williamhoff.tech/windows-and-doors-detection daily 0.75 2026-03-06 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/e01998bd-0c77-417e-9db4-f33f8d6a775b/window-doors-input.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/aceb0ea3-69ee-4d03-97c2-a9925fdceae1/window-doors-lines.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/63ae47e7-f507-4401-89c1-dffa439a1660/window-doors-vanishing.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/67b5dca0-6544-4eeb-9b4f-2d550abdb944/window-doors-rectified.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/263eea8a-f61f-4ecb-ae6e-4e89bd7e2823/window-doors-detections.png https://www.williamhoff.tech/tire-and-defect-detection daily 0.75 2026-03-06 https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/4982bb40-2e84-4f3c-be6a-883aea0cf285/tire-positions.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/33ee48e7-f301-4cb1-9c9b-e789ca2b2701/tire-Blender.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/6df34760-c502-4d7a-a0ee-b200aa8c529f/tire-synth1.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/9e0ffd34-af9c-417e-9d88-1f5a16863c82/tire-synth2.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/916fd476-a7b2-4200-a370-f3779e8086fc/tire-result1.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/03c9dd9e-09a8-4233-8bd0-59acdaeec336/tire-result2.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/4684374b-9ccb-445f-96ad-640fc1a3bc81/tire-result3.png https://images.squarespace-cdn.com/content/v1/5fb9c8b175e654100b273fd3/a240e2d2-6b94-4e4e-bed8-70533c80a8c8/tire-result4.png