Inspiration
Buying secondhand products often requires buyers to trust carefully selected photos and subjective descriptions. Sellers may unintentionally—or intentionally—leave out details that matter, making it difficult to judge an item's true condition before purchasing.
I wanted to make secondhand shopping more transparent by giving buyers a better way to inspect products before committing to a purchase.
What it does
RealResale helps buyers make more informed secondhand purchases by providing richer product information. In addition to conventional listings, buyers can inspect interactive 3D models, view detailed images, and examine visible flaws from different angles. Our goal is to help buyers understand what they are purchasing, reduce uncertainty, and shop with greater confidence.
How I built it
Used Copilot. I built the MVP as a locally running web application using React, JavaScript, and Vite for the frontend, with Node.js and Express for the backend.
- 3D reconstruction: I use CUDA-enabled COLMAP 4.2.1 to reconstruct 3D geometry from photographs on a local PC, then generate a mesh using Delaunay triangulation.
- 3D visualization: Three.js renders the reconstructed PLY models in the browser, allowing users to rotate and zoom in on products.
- Image and video processing: Browser APIs, including
MediaRecorderand Canvas, capture and extract video frames. The server uses Sharp for image validation and processing. - Backend and storage: An Express-based REST API handles scanning, listings, status checks, and delivery of models and images. SQLite stores application data, while files are stored on the local disk.
The MVP runs locally without cloud services or external AI models.
Challenges I ran into
Reconstructing a useful 3D model from ordinary photographs is challenging. Image quality, camera movement, lighting, reflective surfaces, and insufficient coverage can all affect reconstruction quality.
Another challenge is making 3D models useful for evaluating secondhand products. A model can communicate shape and appearance, but it may not reliably reveal every scratch, internal defect, or functional issue. This is why detailed images remain important alongside 3D visualization.
Accomplishments that we're proud of
We built an end-to-end local MVP that connects photo-based 3D reconstruction with an interactive browser viewer and a backend for listings and product data.
More importantly, we established a technical foundation for a secondhand marketplace centered on helping buyers inspect products—not simply browse seller-created listings.
What I learned
Problem solving with vibe coding
What's next for RealResale
Next, we want to improve reconstruction quality and make the scanning process easier for sellers. We also plan to develop better ways to present visible defects, product dimensions, and condition details so buyers can evaluate listings more confidently.
Ultimately, we want RealResale to make secondhand shopping more transparent by helping buyers see more of what they are actually buying—before they pay for it.
Log in or sign up for Devpost to join the conversation.