README.md

---
tags:
  - deep-learning
  - agriculture
  - vineyards
  - segmentation
  - logits
license: mit
datasets:
  - dataset_vineyardLogits_sigmoid
task_categories:
  - image-segmentation
---
# Vineyard Logits Sigmoid Dataset 🍇

## 📌 Overview
The **dataset_vineyardLogits_sigmoid** is a collection of **logits and labels** used for training and testing deep learning models in **precision agriculture**.  

💡 **Key Details**:
- **Binary classification task** with **one class**.
- **Sigmoid activation function** used to output probabilities.
- **Optimized for distinguishing vine plants from background elements**.

This dataset provides valuable logits from models trained on vineyard segmentation tasks, enabling further research and development in precision agriculture.

---

## 📊 Hyperparameters 

The dataset consists of **three distinct datasets** used for **binary classification**. Below are the key hyperparameters used during training and testing:

1. **Split Ratio**  
   - The dataset is split **80:20** (80% training, 20% testing).  

2. **Learning Rate**  
   - Initial **learning rate: 0.001**.

3. **Batch Sizes**  
   - **Training batch size**: **30**
   - **Testing batch size**: **3**  
   - This ensures efficient model training and evaluation.

---

## 📂 Dataset Structure

```plaintext
dataset_vineyardLogits_sigmoid
├── deeplab_EARLY_FUSION_t1
├── deeplab_EARLY_FUSION_t2
├── deeplab_EARLY_FUSION_t3
├── deeplab_GNDVI_t1
├── deeplab_GNDVI_t2
├── deeplab_GNDVI_t3
├── deeplab_NDVI_t1
├── deeplab_NDVI_t2
├── deeplab_NDVI_t3
├── deeplab_RGB_t1
├── deeplab_RGB_t2
├── deeplab_RGB_t3
├── segnet_EARLY_FUSION_t1
├── segnet_EARLY_FUSION_t2
├── segnet_EARLY_FUSION_t3
├── segnet_GNDVI_t1
├── segnet_GNDVI_t2
├── segnet_GNDVI_t3
├── segnet_NDVI_t1
├── segnet_NDVI_t2
├── segnet_NDVI_t3
├── segnet_RGB_t1
├── segnet_RGB_t2
├── segnet_RGB_t3
└── README.md
```

---

## 📑 Contents

- **model_modality_fold_n/pred_masks_train**: Logits from the training set.
- **model_modality_fold_n/pred_masks_test**: Logits from the test set.

---

## 📸 Data Description

- **Model Logits**  
  The dataset consists of logits generated by **DeepLabV3** and **SegNet** during training and testing. These logits are **unnormalized raw scores** before applying the **sigmoid activation function**.

- **Original Images**  
  The images originate from aerial multispectral imagery collected from **three vineyards in central Portugal**:
  - **Quinta de Baixo (QTA)**
  - **ESAC**
  - **Valdoeiro (VAL)**  

  ✅ **Captured at 240x240 resolution** using:
  - **X7 RGB camera**
  - **MicaSense Altum multispectral sensor**  

  ✅ Includes **RGB and Near-Infrared (NIR) bands**, enabling vegetation indices like **NDVI** and **GNDVI**.  
  ✅ **Ground-truth annotations available** for vineyard segmentation.

📌 **For more details**, refer to the dataset:  
[Cybonic, "DL Vineyard Segmentation Study," v1.0, GitHub, 2024](https://github.com/Cybonic/DL_vineyard_segmentation_study)

---

## 📥 How to Use

### **1️⃣ Load in Python**
To load the dataset directly from Hugging Face:

```python
from datasets import load_dataset
dataset = load_dataset("wilgomoreira/dataset_vineyardLogits_sigmoid")
print(dataset)
```

### **2️⃣ Download Specific Files**
To download a specific file:

```bash
wget https://huggingface.co/datasets/seu-usuario/dataset_vineyardLogits_sigmoid/resolve/main/logits_train.npz
```

---

## 🛠 License
This dataset is released under the **MIT License**.  
Please make sure to comply with the license terms when using this dataset.

---

## 🙌 Acknowledgments
This dataset was created by **Wilgo Cardoso** for research in **precision agriculture and deep learning segmentation**.

---

## 📧 Contact
For any questions or collaborations, please contact:  
✉️ **wilgo.moreira@isr.uc.pt**