What is the PaliGemma JSONL Annotation Format?

Overview

PaliGemma is a multimodal model developed by Google. You can export image datasets in the PaliGemma format for use in fine-tuning models. PaliGemma uses data formatted in a JSONL structure.

Format Description

Below, learn the structure of PaliGemma JSONL.


{"prefix": "detect fracture", "suffix": " fracture", "image": "rot_0_7471_png_jpg.rf.30ec1d3771a6b126e7d5f14ad0b3073b.jpg"}{"prefix": "detect fracture", "suffix": " fracture", "image": "flip_0_5824_png_jpg.rf.abe91e2cd085f0d47e35ef7021ff8549.jpg"}
{"prefix": "detect fracture", "suffix": " fracture", "image": "all_0_8542_png_jpg.rf.6bcad49d206468d7720d727caca95724.jpg"}{"prefix": "detect fracture", "suffix": " fracture", "image": "bri_0_592_png_jpg.rf.8d8630701ed43bb703fdad74c8765b26.jpg"}
{"prefix": "detect fracture", "suffix": " fracture ;  fracture ;  fracture", "image": "all_0_2435_png_jpg.rf.e602a2f82935f4fdba988b280ab11b7e.jpg"}

Split and Merge Datasets

With Roboflow supervision, an open source Python package with utilities for completing computer vision tasks, you can merge and split detections in PaliGemma JSONL. Read our dedicated guides to learn how to merge and split PaliGemma JSONL detections.

Split detections in PaliGemma JSONL

No items found.

Supported Models

Below, see model architectures that require data in the PaliGemma JSONL format when training a new model.

On each page below, you can find links to our guides that show how to plot predictions from the model, and complete other common tasks like detecting small objects with the model.