Automotive V1

Overview

Nyris AEM V1 is a domain-specialized image embedding model developed specifically for automotive part identification. It is trained on diverse automotive component data, including engine parts, bodywork, accessories, and mechanical assemblies.

The model is optimized solely for instance-level retrieval, focusing on fine-grained distinctions commonly found in spare parts, where visually similar components differ by only subtle shape details or manufacturing variations.

Architecture

The Automotive V1 encoder outputs a 768-dimensional embedding, balancing retrieval accuracy with computational efficiency.

Use Case

This model serves as a strong baseline for retrieval in automotive workflows and represents nyris's domain-dependent approach to industrial visual search.

Target Dataset

This model is specifically trained and evaluated on the Intercars dataset, which contains automotive spare parts from a leading European automotive parts distributor.