---
dataset_info:
- config_name: all
  features:
  - name: image
    dtype: image
  - name: depth
    dtype: image
  - name: mask
    dtype: image
  splits:
  - name: train
    num_bytes: 22712394633.0
    num_examples: 12000
  download_size: 22677841215
  dataset_size: 22712394633.0
- config_name: default
  features:
  - name: image
    dtype: image
  - name: depth
    dtype: image
  - name: mask
    dtype: image
  splits:
  - name: train
    num_examples: 12000
  download_size: 21573770746
  dataset_size: 21702540415
configs:
- config_name: all
  data_files:
  - split: train
    path: all/train-*
- config_name: blender
  data_files:
  - split: train
    path: rendered/*.tar.gz
- config_name: sdxl
  data_files:
  - split: train
    path: generated/*.tar.gz
language:
- en
license: gpl-3.0
tags:
- vision
- image segmentation
- instance segmentation
- object detection
- synthetic
- sim-to-real
- depth estimation
- image to image
annotations_creators:
- machine-generated
pretty_name: SynWBM Dataset
size_categories:
- 1K<n<10K
task_categories:
- object-detection
- image-segmentation
- depth-estimation
- image-to-image
task_ids:
- instance-segmentation
- semantic-segmentation
---
# The SynWBM (Synthetic White Button Mushrooms) Dataset!

![Sample example](https://huggingface.co/datasets/ABC-iRobotics/SynWBM/resolve/main/example.png "Sample example")

Synthetic dataset of white button mushrooms (Agaricus bisporus) with instance segmentation masks and depth maps.

### Dataset Summary

The SynWBM Dataset is a collection of synthetic images of white button mushroom. The dataset incorporates rendered (using Blender) and generated (using Stable Diffusion XL) synthetic images for training mushroom segmentation models. Each image is annotated with instance segmentation masks and depth maps. The dataset explicitly marks synthetic samples based on their creation method (either rendered from the Blender scene or generated using Stable Diffusion XL) to facilitate sim-to-real performance tests on different synthetic datasets. The dataset does not provide real-world samples for testing, it only contains the synthetic training data.

### Supported Tasks and Leaderboards

The dataset supports tasks such as semantic segmentation, instance segmentation, object detection, depth estimation, image-to-image, and pre-training models for sim-to-real transfer.

## Dataset Structure

### Data samples

The samples of the dataset are 1024x1024x3 images in PNG format. The annotations are 1024x1024x1 PNG images representing the instance segmentation masks (using uint16 data format) and 1024x1024x1 PNG images representing depth data.

### Data Fields

The data fields are:

1) 'image': 1024x1024x3 PNG image
2) 'depth': 1024x1024x1 PNG image
3) 'mask': 1024x1024x1 unit16 PNG image

⚠️ Note: Instance masks are stored as uint16 PNG images, where each pixel value corresponds to an instance ID. Many default image viewers will render them as black; for visualization, we recommend converting them to RGB with a color palette.

### Data Splits

The dataset contains only a training split for all image collections, but provides three configs (`all`, `blender` and `sdxl`) to load the images generated by different methods.

### Limitations

- The dataset only contains synthetic images; no real-world samples are provided.  
- Stable Diffusion XL generations may contain artifacts or unrealistic structures.  
- Depth annotations are derived from rendering/generation pipelines and may not reflect real-world sensor noise.

## Usage

You can load the dataset using 🤗 Datasets:

```python
from datasets import load_dataset

# Load all synthetic images
ds = load_dataset("ABC-iRobotics/SynWBM", name="all")

# Load only Blender-rendered images
ds_blender = load_dataset("ABC-iRobotics/SynWBM", name="blender")

# Load only SDXL-generated images
ds_sdxl = load_dataset("ABC-iRobotics/SynWBM", name="sdxl")

# Example visualization
example = ds["train"][0]
image, mask, depth = example["image"], example["mask"], example["depth"]
```

### Intended Use

The dataset is intended for research in:
- instance segmentation of mushrooms
- sim-to-real transfer learning
- synthetic-to-real domain adaptation
- benchmarking data generation pipelines

## Dataset Creation

### Curation Rationale

The dataset was created to train instance segmentation models on synthetic data for mushroom instance segmentation in real-world images.

### Source Data

The data is generated using two methods:

- Blender-rendered images are produced from a fully procedural Blender scene, with annotations generated using the [Blender Annotation Tool (BAT)](https://github.com/ABC-iRobotics/blender_annotation_tool)
- Stable Diffusion XL images are generated using ComfyUI with a [custom workflow](https://huggingface.co/datasets/ABC-iRobotics/SynWBM/blob/main/mushroom_worflow.json) provided in this repository.

## Citation Information

Coming Soon

## License
This dataset is released under [GNU General Public License v3.0](https://www.gnu.org/licenses/gpl-3.0.en.html).