Anshulgada/RT-PDS

Real Time Pothole Detection System Training Dataset & Model Files

Model Files

Primary model: pothole-detector.pt — this is the actual pre-trained YOLOv10b model used for this project.

You can download it directly from the Hugging Face Hub:

• Direct download link: pothole-detector.pt

• Python snippet:

from huggingface_hub import hf_hub_download

model_path = hf_hub_download(
    repo_id="Anshulgada/RT-PDS",
    filename="pothole-detector.pt"
)

---

Other Available Ultralytics Variants

Model	Description
yolov10n.pt	Nano model, smallest & fastest
yolov10s.pt	Small model
yolov10m.pt	Medium model
yolov10b.pt	Base model
yolov10l.pt	Large model
yolov10x.pt	Extra large, highest accuracy

By default, these Ultralytics weights are available from:

👉 https://github.com/ultralytics/assets/releases/download/v8.3.0/yolov10{variant-name[n,s,m,b,l,x]}.pt

A backup of these models may also be hosted on Hugging Face Hub.

---

Dataset Structure

The dataset follows the standard YOLO format with separate directories for training, validation, and testing. Each split contains both images/ and labels/ subdirectories with matching filenames.

Yolo/
├── Inference Images/       # Example images for quick testing
└── Datasets/
    ├── train/
    │   ├── images/         # ~38k training images
    │   └── labels/         # YOLO-format labels
    ├── valid/
    │   ├── images/         # 6k validation images
    │   └── labels/         
    └── test/
        ├── images/         # 10k test images
        └── labels/         

You can download it directly from the Hugging Face Hub:

• Direct download link: Yolo.zip

• Python snippet:

from huggingface_hub import hf_hub_download

# Download the zipped YOLO dataset
dataset_path = hf_hub_download(
    repo_id="Anshulgada/RT-PDS",
    filename="Yolo.zip",
    repo_type="dataset"
)

print("Dataset downloaded to:", dataset_path)