This repository provides evaluation datasets, models, and evaluation scripts for Persian image captioning and text-image retrieval, developed as part of the Persian Text-Image Retrieval (PTIR) framework. Evaluation Dataset derived from the COCO dataset, this benchmark includes 117,000 images with Persian captions, tailored for retrieval evaluation. The COCO-PIC dataset is also available in the Hugging Face Datasets.
