monocular-vision

OpenCap Monocular: 3D Human Kinematics and Dynamics From a Single Smartphone Video

DERD-Net: Learning Depth from Event-based Ray Densities (NeurIPS 2025 Spotlight)

Lightweight UAV-SAR perception system for aerial human detection, monocular GPS geolocation, and automated rescue alerting using standard UAV telemetry and edge AI.

A Python library for measurement, observation, and calibrated judgment from camera images — the perception layer for human-helper AI applications.

PX4 SITL visual servoing demo: follow a moving AprilTag with a downward camera, attitude compensation, ROS 2, Gazebo and keyboard teleop

Deepixel develops real-time human understanding algorithms from a single RGB camera, with full in-house capability spanning data acquisition, annotation, model training, and platform-optimized deployment—ensuring robust, efficient, and production-ready vision systems.

iOS ball-tracer app in SwiftUI, and the monocular distance engine behind it — turning a single phone video into ball speed, launch angle and (eventually) carry. Source-available.

2026 东南大学电赛校赛 D题——基于单目视觉的饮料杯测量装置。项目基于 Raspberry Pi 5、IMX219 单目摄像头、OpenCV 和自制 INA219 电流检测电路的饮料杯测量系统，可实现目标距离、杯高、液位、液体颜色与体积测量，并支持相机内外参标定、OLED 显示和实时功耗监测。项目获得校赛一等奖。

High-performance facial landmark detection and tracking library by Deepixel. CPU-only, real-time inference using TensorFlow Lite, OpenCV, and DeepCore. Outputs 106 facial landmarks with head pose estimation and Python API support.

Direct Sparse Odometry

Real-time, CPU-only wrist landmark tracking library by Deepixel. Uses monocular vision to estimate 8-point cylindrical wrist landmarks, left/right wrist pose, and visibility confidence with a Python API.

Single-camera 3D morphology reconstruction from one pseudo-overlapped speckle image using DivideNet V3, SIFT, and calibrated triangulation.