marqo-fashionSigLIP

Name: marqo-fashionSigLIP
Author: Marqo

by Marqo

Fashion-domain visual embedding model fine-tuned with Generalised Contrastive Learning

353Kdl/month

88likes

203Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

Marqo/marqo-fashionSigLIP

Feature URI

mixpeek://image_extractor@v1/marqo_fashionsiglip_v1

Overview

Marqo FashionSigLIP is a ViT-B/16-SigLIP model fine-tuned on over 1M fashion products using Generalised Contrastive Learning (GCL). Unlike generic CLIP models, it trains on rich fashion metadata including categories, styles, colors, materials, and fine-grained product details, delivering up to 57% improvement in MRR and recall over previous fashion-specific models.

On Mixpeek, FashionSigLIP powers domain-specific visual search for e-commerce and retail, where generic embeddings miss style nuances like fabric texture, color palette, and silhouette that are critical for product discovery and recommendation.

Architecture

ViT-B/16-SigLIP (webli) backbone fine-tuned with Generalised Contrastive Learning on fashion-specific metadata (categories, styles, colors, materials, keywords). Sigmoid contrastive loss for efficient pairwise training. 768-dimensional shared image-text embedding space.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so marqo-fashionSigLIP runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The vector name has to match a vector index on the collection.
          vectors: { "image-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// image_extractor@v1 runs google/siglip-base-patch16-224
// (768-d) over a bucket, with no inference of your own.

Capabilities

Fashion-optimized 768-dim visual embeddings
Text-to-image and image-to-image product search
Fine-grained attribute awareness (color, material, style, silhouette)
57% MRR improvement over FashionCLIP 2.0

Use Cases on Mixpeek

Visual product search for fashion e-commerce (find similar garments by style)

Automated product categorization and attribute tagging for catalogs

Recommendation engines that understand fashion-specific visual similarity

Benchmarks

Dataset	Metric	Score	Source
Fashion Product Retrieval	MRR improvement vs FashionCLIP 2.0	+57%	Marqo, 2024: marqo-FashionCLIP GitHub
Fashion Category Classification	Recall improvement vs FashionCLIP 2.0	+57%	Marqo, 2024: marqo-FashionCLIP GitHub