MCD-2.6m

MCD-2.6m is a collection of 2,604,450 agricultural and plant images distributed in 49 Parquet shards. It combines images of multiple crops collected across several institutions and field-imaging projects.

The release contains one train split. Images are embedded in the Parquet files and can be decoded directly with the Hugging Face datasets library.

Dataset Structure

Each example contains exactly three columns:

Column	Type	Description
row_id	int64	The integer row identifier retained from the processed source data. It is intended for record tracking and should not be assumed to be globally unique or contiguous across all sources.
image	datasets.Image	The embedded image bytes. Hugging Face datasets decodes this field into a PIL image by default.
commonName	string	The common name of the crop or plant associated with the image.

Source-dataset identity is documented below but is not included as a row-level column in this compact release.

Loading the Dataset

from datasets import load_dataset

dataset = load_dataset("XIANG-Shuai/MCD-2.6m", split="train")
example = dataset[0]
image = example["image"]
crop_name = example["commonName"]

Citation

If you use MCD-2.6m in your research, please cite the SPROUT paper:

@article{xiang2026sprout,
  title   = {SPROUT: A Scalable Diffusion Foundation Model for Agricultural Vision},
  author  = {Xiang, Shuai and Burridge, James and Liu, Shouyang and Lu, Hao and Fukatsu, Tokihiro and Zheng, Yinqiang and Guo, Wei},
  journal = {arXiv preprint arXiv:2603.27519},
  year    = {2026}
}

Data Sources

The following counts are rounded source-level estimates. They describe the source inventory and should not be summed to recover the exact release size because harmonization and final packaging affect the number of retained records.

Crop	Source	Image count	License
cassava	UTOKYO	48K	CC BY 4.0
grape	UTOKYO	150K	CC BY 4.0
Jerusalem	UTOKYO	73K	CC BY 4.0
maize	HUST	121K	CC BY 4.0
maize	UTOKYO	39K	CC BY 4.0
rice	NJAU	202K	CC BY 4.0
rice	UTOKYO	190K	CC BY 4.0
sorghum	SGxP	544K	CC0 1.0
sorghum	UTOKYO	174K	CC BY 4.0
soybean	UTOKYO	178K	CC BY 4.0
sugarbeet	UTOKYO	48K	CC BY 4.0
tomato	UTOKYO	35K	CC BY 4.0
wheat	GWFSS	1.0M	CC BY-SA 4.0
wheat	UTOKYO	200K	CC BY 4.0

Image Processing

The unlabeled images from the Global Wheat Full Semantic Organ Segmentation Dataset (GWFSS) were converted into 512 x 512 image patches using sliding-window cropping before inclusion in MCD-2.6m. These processed GWFSS images remain subject to the original CC BY-SA 4.0 license and attribution requirements.

The combined data were harmonized into a common Parquet representation. The images are stored as embedded bytes rather than external file paths.

Licensing

MCD-2.6m is a mixed-license collection. The applicable license is determined by the source shown in the table above:

• SGxP data are released under the CC0 1.0 Universal Public Domain Dedication.
• GWFSS data, including the 512 x 512 sliding-window crops, are distributed under CC BY-SA 4.0.
• The HUST, NJAU, and UTOKYO data in this release are distributed under CC BY 4.0.

Users must comply with the license and attribution requirements of each source. The license metadata at the top of this card lists all licenses represented in the collection and does not replace the source-specific terms.

Limitations

The source contribution counts are approximate. Crop labels use common names and may include naming variants. This compact release does not contain a row-level source identifier, scientific taxonomy, segmentation masks, captions, or other annotations beyond commonName.

GWFSS References

• GWFSS publication
• GWFSS dataset record and license