Ground type detection P

📦 Segmentation Dataset PASCAL

This repository contains a dataset labeled with Roboflow in PASCAL VOC format, designed for detecting and segmenting different types of terrain surfaces and obstacle negotiability.

📄 Format

-Export format: XML PASCAL VOC-Number of classes: 10-Number of picture: 3040-Annotation source: Roboflow-Annotation type: Pascal VOC standard

🏷️ Available Classes

| ID | Label       | Description                                |----|-------------|--------------------------------------------|| 0  | `concrete`   | Hard surface, pavement or concrete         | 1  | `dirtroad`   | Dirt road or path                          | 2  | `grass`       | Grass or grassy terrain                    | 3  | `gravel`      | Gravel surface or loose rocky path         | 4  | `NS`           | Not traversable (Non-Surmountable)         | 5  | `S50`         | 50% Surmountable slope                     | 6  | `S75`         | 75% Surmountable slope                     | 7  | `S100`       | Fully Surmountable (100%)                  | 8  | `wood`      | Wooden surface

These classes are meant to support the training of terrain perception models, especially for legged robots operating in unstructured environments.

📁 Dataset Structure

dataset/├── train/ # Training images and labels├── valid/ # Validation images and labels├── test/ # Test images and labels├── README.dataset.txt └── README.roboflow.txt