Benchmark Test Splits used for Evaluating MIRCaps-VLM on the Image Captioning Task
收藏官方服务：
资源简介：
The following benchmark splits are used to evaluate the image captioning performance of MIRCaps-VLM. Each sample consists of an image paired with its corresponding ground-truth captions. COCO (Karpathy test split, 5K images) Flickr30k (Karpathy test split, 1K images) MIRCaps (test split, 7.4K images) NoCaps (validation split, 4.5 images) TextCaps (test split, 3K images)
提供机构：
Zenodo创建时间：
2026-07-14
