Innovatiana's Luxury Fashion Fraud Detection (LFFD)
This open source dataset brings together a curated collection of images of Chanel products, whether authentic or not, in order to support the development of AI systems applied to fashion and luxury, in particular in the fight against counterfeiting. The dataset was structured by Innovatiana, without any affiliation or approval from the Chanel brand.
Several hundreds of high-quality images, JPG format, classified by type of product (clothing, bags, accessories, perfumes...)
Use strictly reserved for research and teaching. The images are the property of their respective rights holders. Innovatiana does not claim any rights to brand visuals
Description
The dataset contains:
- Various images of Chanel products: handbags (ex. 2.55), clothing, shoes, accessories, perfumes
- A variety of online sources (marketplaces, forums, e-commerce, etc.)
- A classification by product categories (each file contains homogeneous visuals)
- Images in format JPG, without annotation or metadata
- No retouching: the images reflect the real conditions of exposure (neutral background, fashion shows, online sales...)
The objective is to facilitate the development of robust visual analysis models, including on images from various contexts.
What is this dataset for?
This dataset is designed to:
- Training counterfeit detection models in luxury fashion
- Recognition of specific products (object detection or classification)
- The analysis of fashion images in an e-commerce or social media context
- Prototyping systems for the detection of visual defects (scratches, deformations, etc.)
- Analysis of the consistency of a product with respect to official references
Can it be enriched or improved?
Yes, several options:
- Annotate images with metadata (authenticity, price, visible defects...)
- Add context information (source, angle, usage...)
- Integrate similar images from other brands to train multi-brand models
- Create an augmented database with detection of logos or patterns specific to Chanel
🔗 Source: Hugging Face — Innovatiana Chanel Dataset for Fraud Detection
Frequently Asked Questions
Does the dataset make it possible to distinguish authentic products from counterfeits?
The dataset contains an unverified mix of images. It can be used to train counterfeit detection models, provided an annotation or validation layer is added.
Can I use this dataset in a commercial project?
No Use is limited to research. Any commercial use would require obtaining the rights to the images concerned.
Why publish this dataset if the images belong to a brand?
The images are taken from visuals available online. The aim is to contribute to research on counterfeiting and to illustrate concrete cases of Computer Vision in luxury. No commercial exploitation is allowed.
