food-ai-nexus/microcolony-domain-adaptation

---
license: cc-by-4.0
task_categories:
- image-classification
tags:
- food-safety
- microscopy
- bacteria
- domain-adaptation
- few-shot-learning
language:
- en
size_categories:
- 1K<n<10K
pretty_name: Microcolony Domain Adaptation (Foodborne Bacteria)
configs:
- config_name: default
data_files:
- split: train
path: train/**
- split: train_fewshot
path: train_fewshot/**
- split: test_standard
path: test_standard/**
- split: test_ood
path: test_ood/**
---

**Microcolony Domain Adaptation (Foodborne Bacteria)** is a microscopy image dataset for foodborne bacterial classification under varying imaging conditions. It was created to support research in adversarial domain adaptation, enabling models trained on standard phase contrast microscopy images to generalize across different optical configurations and biological conditions.

This dataset accompanies the publication: *Bhattacharya, S., Wasit, A., Earles, M., Nitin, N., & Yi, J. (2025). Enhancing AI microscopy for foodborne bacterial classification using adversarial domain adaptation to address optical and biological variability. Frontiers in Artificial Intelligence, 8, 1632344.* [doi: 10.3389/frai.2025.1632344](https://doi.org/10.3389/frai.2025.1632344)

# Content

The dataset contains microscopy images of six foodborne bacterial species imaged under a source domain (standard conditions) and multiple target domains (varying optical and biological conditions). It is structured into four splits to support both standard supervised training and domain adaptation experiments.

| Split | Description | Images |
| --- | --- | --- |
| `train` | Standard phase contrast images (60x, 3h), with class subdirectories | 539 |
| `train_fewshot` | Small labeled samples from target domains for few-shot adaptation | 150 |
| `test_standard` | Held-out standard phase contrast images (same domain as `train`) | 90 |
| `test_ood` | Out-of-distribution images under varying imaging conditions | 420 |

# Classes

| Code | Species | Full Name |
| --- | --- | --- |
| `Bc` | *Bacillus coagulans* | Gram-positive spore-forming bacterium |
| `Bs` | *Bacillus subtilis* | Gram-positive model organism |
| `Ec` | *Escherichia coli* 1612 | Gram-negative foodborne pathogen |
| `Li` | *Listeria innocua* | Non-pathogenic *Listeria* surrogate |
| `SE` | *Salmonella enterica* Enteritidis | Foodborne pathogen |
| `ST` | *Salmonella enterica* Typhimurium | Foodborne pathogen |

# Imaging Conditions

| Domain | Objective | Incubation | Modality | Split |
| --- | --- | --- | --- | --- |
| `phase_contrast_60x_3h` | 60x | 3 h | Phase contrast | `train`, `train_fewshot`, `test_standard` |
| `20x-3h` | 20x | 3 h | Phase contrast | `train_fewshot`, `test_ood` |
| `20x-5h` | 20x | 5 h | Phase contrast | `train_fewshot`, `test_ood` |
| `20x` | 20x | 3 h | Phase contrast | `test_ood` |
| `brightfield` | 60x | 3 h | Brightfield | `train_fewshot`, `test_ood` |
| `defocus` | 60x | 3 h | Phase contrast (defocused) | `train_fewshot`, `test_ood` |
| `agar15` | 60x | 3 h | Phase contrast (1.5% agar) | `test_ood` |

# Uses

This dataset is intended for:
- **Image classification** of foodborne bacterial microcolonies.
- **Domain adaptation** research, where models trained on the source domain (`phase_contrast_60x_3h`) are evaluated on target domains.
- **Few-shot learning** experiments using the `train_fewshot` split.

```python
from datasets import load_dataset

# Load all splits
ds = load_dataset("food-ai-nexus/microcolony-domain-adaptation")

# Load only the standard train/test splits
train = ds["train"]
test_std = ds["test_standard"]
test_ood = ds["test_ood"]
```

# License

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) License.

# Citation

```bibtex
@article{bhattacharya2025microcolony,
title = {Enhancing AI microscopy for foodborne bacterial classification using adversarial domain adaptation to address optical and biological variability},
author = {Bhattacharya, Siddhartha and Wasit, Aarham and Earles, J. Mason and Nitin, Nitin and Yi, Jiyoon},
journal = {Frontiers in Artificial Intelligence},
volume = {8},
pages = {1632344},
year = {2025},
doi = {10.3389/frai.2025.1632344}
}
```

# Source

Original dataset: [Zenodo 10.5281/zenodo.16741157](https://zenodo.org/records/16741157)
Code repository: [GitHub food-ai-engineering-lab/microcolony-domain-adaptation-frai](https://github.com/food-ai-engineering-lab/microcolony-domain-adaptation-frai)

---
许可证：CC BY 4.0
任务类别：
- 图像分类（Image Classification）
标签：
- 食品安全（Food Safety）
- 显微成像（Microscopy）
- 细菌（Bacteria）
- 域自适应（Domain Adaptation）
- 少样本学习（Few-shot Learning）
语言：
- 英语
规模类别：
- 1K<n<10K
漂亮名称：微菌落域自适应（食源性细菌）
配置项：
- 配置名称：default
数据文件：
- 拆分：train
路径：train/**
- 拆分：train_fewshot
路径：train_fewshot/**
- 拆分：test_standard
路径：test_standard/**
- 拆分：test_ood
路径：test_ood/**
---

**微菌落域自适应（食源性细菌）**是一款面向不同成像条件下食源性细菌分类任务的显微图像数据集。本数据集旨在支撑对抗域自适应（Adversarial Domain Adaptation）相关研究，使在标准相差显微成像图像上训练的模型能够泛化至不同光学配置与生物条件的场景。

本数据集配套发表论文：*Bhattacharya, S., Wasit, A., Earles, M., Nitin, N., & Yi, J. (2025). 利用对抗域自适应解决光学与生物变异性，提升食源性细菌分类的AI显微技术. 《人工智能前沿》, 8, 1632344.* [doi: 10.3389/frai.2025.1632344](https://doi.org/10.3389/frai.2025.1632344)

## 数据集内容
本数据集包含6种食源性细菌的显微图像，这些图像分别采集自源域（标准成像条件）与多个目标域（不同光学与生物条件）。数据集划分为四个拆分，以支持标准监督训练与域自适应实验。

| 拆分名称 | 描述 | 图像数量 |
| --- | --- | --- |
| `train` | 标准相差显微图像（60倍物镜，孵育3小时），含按类别划分的子目录 | 539 |
| `train_fewshot` | 来自目标域的少量标注样本，用于少样本自适应 | 150 |
| `test_standard` | 预留的标准相差显微图像（与`train`同域） | 90 |
| `test_ood` | 不同成像条件下的分布外（Out-of-distribution）图像 | 420 |

## 类别列表
| 类别代码 | 物种简称 | 完整学名与说明 |
| --- | --- | --- |
| `Bc` | *Bacillus coagulans* | 革兰氏阳性（Gram-positive）产芽孢细菌 |
| `Bs` | *Bacillus subtilis* | 革兰氏阳性（Gram-positive）模式生物 |
| `Ec` | *Escherichia coli* 1612 | 革兰氏阴性（Gram-negative）食源性致病菌 |
| `Li` | *Listeria innocua* | 非致病性李斯特菌替代株 |
| `SE` | *Salmonella enterica* Enteritidis | 食源性致病菌 |
| `ST` | *Salmonella enterica* Typhimurium | 食源性致病菌 |

## 成像条件
| 域名称 | 物镜倍率 | 孵育时长 | 成像模态 | 适用拆分 |
| --- | --- | --- | --- | --- |
| `phase_contrast_60x_3h` | 60x | 3小时 | 相差显微成像（Phase Contrast） | `train`、`train_fewshot`、`test_standard` |
| `20x-3h` | 20x | 3小时 | 相差显微成像（Phase Contrast） | `train_fewshot`、`test_ood` |
| `20x-5h` | 20x | 5小时 | 相差显微成像（Phase Contrast） | `train_fewshot`、`test_ood` |
| `20x` | 20x | 3小时 | 相差显微成像（Phase Contrast） | `test_ood` |
| `brightfield` | 60x | 3小时 | 明场成像（Brightfield） | `train_fewshot`、`test_ood` |
| `defocus` | 60x | 3小时 | 离焦相差显微成像（Defocused Phase Contrast） | `train_fewshot`、`test_ood` |
| `agar15` | 60x | 3小时 | 相差显微成像（1.5%琼脂培养基） | `test_ood` |

## 应用场景
本数据集适用于：
- **食源性细菌微菌落图像分类**任务
- **域自适应**研究：将在源域（`phase_contrast_60x_3h`）上训练的模型在目标域上进行评估
- 基于`train_fewshot`拆分的**少样本学习**实验

python
from datasets import load_dataset

# 加载所有拆分
ds = load_dataset("food-ai-nexus/microcolony-domain-adaptation")

# 仅加载标准训练与测试拆分
train = ds["train"]
test_std = ds["test_standard"]
test_ood = ds["test_ood"]

## 许可证
本数据集采用知识共享署名4.0国际许可协议（CC BY 4.0）进行授权。

## 引用格式
bibtex
@article{bhattacharya2025microcolony,
title = {"Enhancing AI microscopy for foodborne bacterial classification using adversarial domain adaptation to address optical and biological variability"},
author = {Bhattacharya, Siddhartha and Wasit, Aarham and Earles, J. Mason and Nitin, Nitin and Yi, Jiyoon},
journal = {Frontiers in Artificial Intelligence},
volume = {8},
pages = {1632344},
year = {2025},
doi = {10.3389/frai.2025.1632344}
}

## 来源
原始数据集：[Zenodo 10.5281/zenodo.16741157](https://zenodo.org/records/16741157)
代码仓库：[GitHub food-ai-engineering-lab/microcolony-domain-adaptation-frai](https://github.com/food-ai-engineering-lab/microcolony-domain-adaptation-frai)