Lense

Overview

Lense is an Android application for selling pre-owned products. You upload a product image. The app classifies the item. The app generates a description. The app estimates the value of the item. The app uses on-device machine learning models.

Problem

To list a pre-owned product for sale, you must write a description and set a price. These tasks take time and involve guesswork. Sellers often do not know the fair market value of their items. Writing good descriptions is tedious.

Approach

Lense uses two on-device AI models to automate this process:

• EfficientNet (TensorFlow Lite 4) classifies the product category from the uploaded image.
• GPT-2 Image Captioning generates a descriptive caption for the listing.

Both models run on-device via HuggingFace. They handle image processing without a server round-trip.

How It Works

1. You upload a product image.
2. EfficientNet classifies the product.
3. GPT-2 Image Captioning generates a description.
4. The app returns the estimated value and a description that is ready to post.

Tech Stack