Feng-GUI vs OpenGazer

Feng-GUI is a visual insights platform that predicts eye-tracking patterns and analyzes attention with AI. It can process websites, advertisements, videos, print materials, shelf layouts, signage and packaging. Reports include heatmaps, gaze plots, focus maps, Areas of Interest and aesthetics analyses, while an AI Assistant turns insights and recommendations into concise summaries. Feng-GUI can detect faces and analyze eight expressions: Neutral, Happy, Surprise, Sad, Anger, Disgust, Fear and Contempt. Report data exports as raw images, JSON or CSV. The service is available through a web browser and an open Platform API, with Model Context Protocol support for AI agents. The company states that MIT benchmarking found results 92% similar to real eye-tracking. It also states compliance with GDPR and PCI DSS and describes itself as a certified Microsoft supplier meeting SSPA and Supplier Data Protection Requirements. Uploaded files and reports are automatically deleted after three months for monthly subscribers and one year for annual subscribers. The documented maximum image size is 20 MB. Business costs $67 a month, billed annually; Basic costs $99 a month, billed monthly.

OpenGazer is a Linux application that uses an ordinary webcam to estimate gaze direction and send it to other programs. It also tracks head position and rotation, and displays the estimated gaze point, head-rotation lines, and calibration target. Setup calls for a camera at 640×480, placed centrally below the monitor. Users calibrate by looking at indicated screen points while keeping their head still, then choose facial tracking points; the project calls for at least four and recommends seven or eight. Gaze estimates appear on screen, print to standard output, and are sent through a local UDP socket on port 20230. Dasher is an example of software that can use the gaze direction for writing with the eyes. OpenGazer is free under GPLv2 and uses the X Window System interface. It is marked Alpha, with a last update listed as 2013-04-24. Building it requires VXL, OpenCV, gtkmm, cairomm, and Boost.