RentSafe AI

Inspiration

Renting a home often begins with a simple problem that becomes expensive later: no one has a reliable record of what the property looked like when the tenant moved in.

A small scratch, stain, crack, broken fixture, or damaged surface can later become a dispute over the security deposit. Photos and videos can provide evidence, but manually comparing two inspections is time-consuming, inconsistent, and difficult when the camera angle, lighting, or furniture placement changes.

We wanted to build a system that turns a normal phone walkthrough into structured, comparable evidence.

What RentSafe AI Does

RentSafe AI is an AI-powered rental inspection system that compares a property's condition at move-in and move-out.

A tenant records a walkthrough of the property. RentSafe AI analyzes the captured images, identifies important areas and visible damage, and creates a structured baseline of the property's condition.

During move-out, the tenant performs another walkthrough. The system aligns the new images with the original inspection and looks for meaningful visual changes.

The workflow is:

Capture → Align → Compare → Detect Changes → Confidence Score → Human Review → Evidence Report

Instead of pretending that every visual difference is automatically "new damage," the system provides a confidence score and highlights uncertain cases for human review.

How We Built It

The core of RentSafe AI combines computer vision with structured evidence generation.

We use image processing and visual feature matching to associate corresponding areas between two inspections, even when the camera position or lighting is not identical. We then analyze visual differences to identify possible changes such as:

• new wall damage
• stains or marks
• broken fixtures
• damaged surfaces
• missing or changed items

The results are organized into a room-by-room inspection report containing the original evidence, the new evidence, detected changes, and confidence levels.

The goal is not simply to say "damage detected." The goal is to answer:

What changed, where did it change, and how confident are we?

What We Learned

The biggest lesson was that real-world computer vision is much harder than comparing two identical images.

A room can look completely different because of lighting, shadows, camera position, furniture movement, reflections, or perspective. A system that works on clean benchmark images can fail badly in an actual apartment.

This pushed us toward a more realistic approach: image alignment + change detection + confidence estimation + human verification, instead of relying on a single binary classifier.

We also learned that useful AI products need to communicate uncertainty rather than hide it.

Challenges We Faced

The hardest challenge was distinguishing real property changes from harmless visual differences.

For example, a brighter light can make an old stain appear new, while moving a chair can reveal a mark that was already present. Perspective changes can also make the same object appear significantly different.

Another challenge was making the system useful without requiring specialized hardware. RentSafe AI is designed around ordinary smartphone images rather than expensive inspection equipment.

The project therefore focuses on creating a practical evidence workflow that is accessible, explainable, and robust enough for real rental inspections.

Why It Matters

RentSafe AI is designed to shift rental disputes from:

"The tenant says it was already there."

to:

"Here is the documented condition, here is the corresponding evidence, and here is what changed."

We believe AI should not replace human judgment in uncertain cases. It should make the evidence easier to collect, compare, understand, and act on.

RentSafe AI — Capture the condition. Protect your deposit.

Built With

• ai
• change
• computer
• css
• deep
• detection
• docker
• fastapi
• image
• learning
• machine
• object
• ocr
• opencv
• postgresql
• processing
• python
• pytorch
• react
• segmentation
• similarity
• tailwind
• typescript
• vision