FENG-GUI ALTERNATIVE

The pioneer drew the first maps. The science redrew them.

TL;DR

Feng-GUI has been simulating human attention since around 2007 — a genuine pioneer, built on classic saliency algorithms: contrast, edges, color, handcrafted rules from the pre-deep-learning era. Heatpoints runs UNISAL, a peer-reviewed deep network (ECCV 2020, open source) trained on real eye-tracking datasets — SALICON, MIT1003, CAT2000 — and wraps it in a full workflow: AI audit, Words Map, Fix + re-score, weekly monitoring. Same question, two technological eras.

TWO ERAS OF THE SAME IDEA

saliency terminal — circa 200780×25

> edge_detection ....... OK

> color_contrast ....... OK

> intensity_filters ....... OK

> center_bias ....... OK

> face_heuristics ....... OK

> render: heatmap [watermarked demo]

> _

HANDCRAFTED RULES — A MARVEL IN ITS DAY
a network trained on real eyes

UNISAL — peer-reviewed at ECCV 2020, open source

SALICONMIT1003CAT2000

Trained on real eye-tracking datasets — faces, text, context.
Then: audit, Words Map, Fix, monitoring.

LEARNED FROM HUMAN GAZE — PEER-REVIEWED IN 2020

Feature by feature

CRITERIONFENG-GUIHEATPOINTS
ApproachClassic saliency algorithms — handcrafted rulesDeep network trained on real eye-tracking data
Scientific basisPre-deep-learning heuristics (~2007 era)UNISAL — ECCV 2020, peer-reviewed, open source
Training dataNone — rules, not trainingSALICON · MIT1003 · CAT2000 — real gaze datasets
Free tierWatermarked demo10 free scans/month + a scanner with no account
AI-written auditIncluded in every report
Word-level copy analysisWords Map — unique to Heatpoints
Fix loopFix + re-score in one click
Live-page monitoringWeekly re-scans
White-label PDF + APISolo, 59 €/mo
Tracking on your visitorsNoneNone — zero script, zero cookie, zero banner
Entry pricePaid plans on feng-gui.comFree 10/mo · Pro 29 € · Solo 59 € · Agency 99 €

Honest note: Feng-GUI deserves respect — it kept algorithmic attention analysis alive for nearly two decades, long before deep learning made it mainstream. If you want a quick, simple check on print, billboards or packaging, its classic saliency pass still does a job. We compete on the science and what surrounds it: a peer-reviewed model trained on real gaze data, plus the audit, copy analysis, fixes and monitoring your web pages actually need.

FAQ

What is Feng-GUI exactly?

One of the true pioneers of algorithmic attention analysis, active since around 2007. It simulates human vision with classic saliency algorithms — contrast, edges, color intensity, handcrafted rules — and renders attention heatmaps for images and pages. The free demo output is watermarked.

What's the difference between classic saliency and deep-learning saliency?

Classic saliency (Feng-GUI's era) computes attention from handcrafted filters: where contrast is high, where edges meet, where colors pop. Deep-learning saliency — UNISAL, the model behind Heatpoints — is a neural network trained on real eye-tracking datasets (SALICON, MIT1003, CAT2000), so it learned from where humans actually looked: faces, text and context included. It was peer-reviewed at ECCV 2020 and is open source.

Is Feng-GUI still worth using in 2026?

For some workflows, honestly, yes. Its simplicity is a feature, and for quick checks on print, billboards or packaging, a classic saliency pass can be enough. For web pages — where copy, hierarchy and CTAs decide conversions — a model trained on real gaze data plus an actionable audit will take you further.

What does Heatpoints add beyond the heatmap?

The workflow around it: an AI-written audit, Words Map (word-level attention on your copy — nobody else does this), one-click Fix with re-score, weekly monitoring of live pages, a predictive MCP for AI agents, and white-label PDF reports. And our N=93 eye-tracking study found 81% of attention lands outside the hero — the heatmap is where analysis starts, not where it ends.

2007 asked the right question. 2026 answers it better.

Scan any page free — 30 seconds, no account, no watermark.