Models Generic General V5.1
nyris

General V5.1

Broad-coverage visual embedding model for generic product identification.

nyris Proprietary Vision Only
Generic
Model Type
#1
Overall Rank
57.6%
Avg P@1
50.3%
Avg mAP@10
768
Embed Dim
336
Input Res
8
Datasets

About This Model

Overview

Nyris GEM V5.1 is a broad-coverage visual embedding model designed for generic product identification. Unlike Automotive V1, which is domain-specific, General V5.1 is trained on a wide distribution of parts and products spanning manufacturing, DIY, and retail domains.

The model is optimized for image-to-image instance retrieval, offering strong fine-grained discrimination without reliance on text or multimodal alignment.

Architecture

With a 768-dimensional embedding, the model is engineered for scalable deployment in real-time retrieval systems.

Capabilities

General V5.1 serves as nyris's most versatile specialist model and demonstrates retrieval behavior tuned specifically for industrial conditions such as:

  • Varying lighting conditions
  • Cluttered backgrounds
  • Catalog-to-real domain shifts

Use Case

This model is the recommended baseline for general-purpose visual product search across diverse domains including manufacturing, DIY, furniture, and retail applications.

Performance Across Datasets

Dataset Category P@1 P@5 R@1 R@5 mAP@10
VPRC 2023 Mixed Retail 41.14% 17.79% 27.68% 51.78% 44.37%
Intercars Automotive 28.18% 22.51% 9.94% 25.11% 29.19%
Stanford Online Products E-commerce 86.87% 64.03% 21.97% 61.80% 70.84%
IKEA Furniture 65.77% 45.44% 19.23% 50.23% 51.68%
Hornbach Hardware/DIY 28.97% 9.96% 28.97% 49.80% 35.99%
ARaymond Industrial 63.38% 56.66% 3.96% 17.70% 44.75%
Products-10K E-commerce 77.09% 61.18% 16.40% 63.64% 64.25%
TOPEX Industrial 69.67% 67.93% 2.18% 10.61% 61.66%
ILIAS Instance Retrieval 69.89% 34.63% 17.12% 35.80% 36.00%
Average 57.63% 43.19% 16.29% 41.33% 50.34%