PGFID: A Labeled Image Dataset of Native Persian Gulf Fish Families for Machine Learning and Ecological Research

Accurate identification of fish families is fundamental to biodiversity assessment, sustainable fisheries management, and ecological monitoring; however, conventional taxonomic workflows remain time-consuming, expert-dependent, and difficult to scale. We present the Persian Gulf Fish Image Dataset (PGFID), a curated collection of 1,350 labeled images spanning 35 distinct fish families native to the Persian Gulf. The dataset includes 942 training images and 408 validation images, all captured in Bushehr (Iran) fish markets between December 2024 and June 2025 using two smartphones (iPhone 13 Pro Max and Samsung Galaxy A05s) under varying backgrounds, lighting conditions, and poses to reflect realistic field environments. Local market names provided by experienced vendors were mapped to scientific family-level labels through cross-referencing taxonomic resources such as FishBase. PGFID serves as a valuable resource for advancing image-based classification, ecological monitoring, and intelligent fisheries systems, supporting research communities working at the intersection of artificial intelligence, marine biology, and biodiversity conservation.