End-to-end pipeline for adapting CLIP-style multi-modal models to domain-specific vision tasks. Uses Triplet Loss and custom retrieval metrics to enable semantic search over image/text pairs, with a focus on fashion datasets.
FashionCLIP
Pipeline de deep learning para el fine-tuning de modelos multi-modales de visión sobre datasets propios. Implementación de Triplet Loss y métricas personalizadas para búsqueda semántica.
PyTorchMulti-modal TransformersFine-TuningModel Architecture