omnigen2
Datasets
All datasets matching “omnigen2”OmniContext
Introduction
As part of OmniGen2, we introduce a new benchmark for in-context generation, OmniContext, which aims to provide a more comprehensive evaluation of models' in-context generation abilities. It incorporates a diverse set of input images and instructions, and utilizes GPT-4.1 for interpretable, metric-driven assessment.
Project Page: https://vectorspacelab.github.io/OmniGen2
Github Repo for OmniContext: https://github.com/VectorSpaceLab/OmniGen2… See the full description on the dataset page: https://huggingface.co/datasets/OmniGen2/OmniContext.X2I2
X2I2 Dataset
2025-08-17: jsons/inpaint_edit/ and images/inpaint_edit/edit_pf_one/ are being fixed, please do not download.
2025-07-15: jsons/reflect/reflect.jsonl has been fixed and updated.
2025-07-05: X2I2 are available now.
X2I2-video-editing
# meta file (en): jsons/video_edit/edit_mv.jsonl
# meta file (zh): jsons/video_edit/edit_mv_zh.jsonl
# images:
cd images/video_edit/edit_mv_0 && cat edit_mv_0.tar.gz.part_* > edit_mv_0.tar.gz && tar… See the full description on the dataset page: https://huggingface.co/datasets/OmniGen2/X2I2.omnigen2-azimuth-ladder-anny-20260901
omnigen2-azimuth-ladder-anny-20260901
An image-edit ladder in the EditScore dataset shape: a candidate measured against a baseline
on the same prompts, one row per (source, edited, instruction) with the per-pair
measurement beside the images. The baseline is OmniGen2; the candidate is the same model
after a camera-control LoRA.
This ladder has no EditScore score. The runs measured recovered azimuth — where the
body actually faces in the generated view — not EditScore's pf / sc /… See the full description on the dataset page: https://huggingface.co/datasets/chibifire/omnigen2-azimuth-ladder-anny-20260901.umm-omnigen2-sftrung0-omnigen2-anny-depth-train
rung0-omnigen2-anny-depth
Rung 0 smoke invocation of the depth-conditioned generation pipeline: one
ANNY render's depth map conditioning one OmniGen2 image. A pipeline record,
not a training corpus.
Five zstd parquet tables in Essential Tuple Normal Form sharing primary
keys: run (one row per invocation), run_params and prompt (satellites),
render (camera and framing of the depth control), asset (image bytes:
the depth control and the generated output).
Generator:… See the full description on the dataset page: https://huggingface.co/datasets/chibifire/rung0-omnigen2-anny-depth-train.genshin_woman_Formal_Outfit_omnigen2_extracted