Honestly, Iβd steer clear of pixel-level comparison or pHash for face verification itβs going to be a support nightmare because even a slight change in lighting or a head tilt will trigger a false rejection. Since you want to keep things lightweight and AI-free, your best bet is to treat the selfie as a "visual receipt" for human audit rather than an automated lock; just force a live camera capture (no gallery uploads) and let the threat of a manager spot-checking the logs act as the deterrent. Also, definitely keep an eye on that 10β20m geofence it's incredibly tight for real-world GPS drift, especially indoors, so you might want to allow a small buffer or flag "near-misses" for review instead of flat-out rejecting them.
Apr 1, 2026
7
Level 20
Laravel Smart Attendance System with Live Geofencing (10β20 Meter Radius) β Feedback Needed
Iβm building a lightweight Laravel-based attendance system where users can check in/out only when they are physically present within a strict 10β20 meter radius of the workplace using live GPS validation.
- βοΈ Core Features Implemented 2.π Real-time geofencing (strict 10β20m radius enforcement)
- π― GPS accuracy filtering (reject low-accuracy readings)
- β±οΈ Live check-in only (no delayed or backdated submissions)
- π« Mock location detection (Android-based checks)
- π± Device tracking (basic fingerprinting for consistency)
- π€³ Selfie Verification Approach (No AI)
To keep the system lightweight and privacy-friendly, Iβm intentionally avoiding AI/ML solutions.
Current approach:
- Capture a selfie during check-in
- Compare it with a previously stored reference image
- Use pixel-level comparison techniques, such as:
- Image difference
- Hashing (e.g., perceptual hash)
- Similarity scoring
β Feedback Iβm Looking For
Iβd really appreciate insights from experienced developers on the following:
- Reliability How reliable is pixel-based image comparison in real-world usage?
- Real-world Variations How should I handle: 1. lighting differences 2. face angle changes 3. different camera quality
- Alternatives (Without AI) Are there any better non-AI approaches for basic face verification?
- Feasibility Is this approach fundamentally flawed for identity verification?
- Improvements Would capturing multiple selfies (2β3 per check-in) improve accuracy or just add noise? π― Goal
The aim is to build a system that is:
- β Simple to integrate
- β‘ Lightweight (no heavy dependencies)
- π Privacy-friendly (no AI processing)
- π‘οΈ Reasonably resistant to misuse
π Closing
Any suggestions, critiques, or alternative approaches would be greatly appreciated. Looking forward to learning from your experience!
Please or to participate in this conversation.