yzwang/X2I-computer-vision

X2I Dataset

• Project Page: https://vectorspacelab.github.io/OmniGen/
• Github: https://github.com/VectorSpaceLab/OmniGen
• Paper: https://arxiv.org/abs/2409.11340
• Model: https://huggingface.co/Shitao/OmniGen-v1

To achieve robust multi-task processing capabilities, it is essential to train the OmniGen on large-scale and diverse datasets. However, in the field of unified image generation, a readily available dataset has yet to emerge. For this reason, we have curated a large-scale unified image generation dataset with unified format for the first time, which we refer to as the X2I dataset, meaning "anything to image".

Task	Datastet
Multi-modal Instruction	X2I-mm-instruction
Subject-driven Editing	X2I-subject-driven
In-context Learning	X2I-in-context-learning
Computer Vision	X2I-computer-vision
Text to Image Generation	X2I-text-to-image

X2I-computer-vision

• ADE

A image segementation dataset with 297,472 samples.

## meta file: ade.jsonl
cd ade
tar -xzvf ade.tar.gz

Download all tar parts (ade/seg_imgs.tar.gz.*) from here

cd ade
cat seg_imgs.tar.gz.* | tar -xzvf -

• Rain100H

A image derain dataset with 100 samples.

## meta file: derain_rain100h.jsonl
cd derain_rain100h
tar -xzvf derain_rain100h.tar.gz

• LOL

A low-light image enhancement dataset with 485 samples.

## meta file: enhance_lol.jsonl
cd enhance_lol
tar -xzvf enhance_lol.tar.gz

• MuLan

A multi-layer image editing dataset with 32,066 samples.

## meta file: mulan.jsonl
cd mulan
tar -xzvf mulan.tar.gz

cd coco2017
tar -xzvf coco2017.tar.gz

• ReasonSeg

A image reasoning segmentation dataset with 1,218 samples.

## meta file: reasonseg.jsonl
cd reasonseg
tar -xzvf reasonseg.tar.gz

• WeatherStream & GoPro

A image derain, desnow and deblur dataset with 3,386 samples.

## meta file: weather-stream_gopro.jsonl
cd weather-stream/rain
tar -zxvf rain.tar.gz

cd weather-stream/snow
tar -zxvf snow.tar.gz

cd gopro
tar -zxvf gopro.tar.gz

cd raindrop
tar -zxvf raindrop.tar.gz

• MultiGen