mkxdxd/carla-dataset-ped

CARLA Stage 2 Pedestrian Dataset

A large-scale driving dataset (Stage 2) focused on pedestrians, captured from CARLA simulator. This dataset contains RGB images and depth maps with camera parameters, specifically tracking pedestrian movements.

Tar File Structure

This dataset is stored in WebDataset format for efficient streaming and loading.

Sharding Strategy

• 2 scenes per shard: Each .tar file contains exactly 2 complete scenes
• 200 frames per scene: Each scene is a video sequence of 200 consecutive frames
• ~400 frames per tar: 2 scenes × 200 frames = 400 frames per shard

Repository Structure

carla-dataset-ped/
├── Town01/
│   └── pedestrian/
│       ├── carla-stage2-000000.tar  ← Scenes 1-2
│       ├── carla-stage2-000001.tar  ← Scenes 3-4
│       └── ...
├── Town02/
│   └── ...
├── Town03/
│   └── ...
├── Town04/
│   └── ...
├── Town05/
│   └── ...
└── Town06/
    └── ...

Shard Contents (WebDataset format)

Each tar file contains samples with the following files per frame:

{scene_id}_{frame_idx:03d}.rgb.png        ← RGB image (1280×704)
{scene_id}_{frame_idx:03d}.depth.npy      ← Depth map (numpy array, 704×1280)
{scene_id}_{frame_idx:03d}.camera.json    ← Camera parameters + matched reference frames
{scene_id}_{frame_idx:03d}.metadata.json  ← Scene info (scene_id, frame_id, town, actor_type)

Key Format: {scene_id}_{frame_idx:03d} (e.g., scene_001_000, scene_001_001, ..., scene_001_199)

Data Format

Each sample contains:

• rgb: PIL.Image (1280×704) - RGB image
• depth: np.ndarray (704, 1280) - Depth map
• camera: dict containing:
  • intrinsic: Camera intrinsic matrix
  • extrinsic: Camera extrinsic matrix
  • matched_references: List of reference image IDs accumulated along the trajectory
• metadata: dict containing:
  • scene_id: Unique scene identifier
  • frame_id: Frame index within the scene (0-199)
  • town: CARLA town name (Town01-Town06)
  • actor_type: Type of actor being followed ("pedestrian")

---

Camera JSON with Reference Frame Mapping

Each camera.json contains camera parameters and matched reference frames:

{
  "intrinsic": { ... },
  "extrinsic": { ... },
  "carla_transform": { ... },
  "matched_references": ["subset_0/0001", "subset_0/0042", "subset_0/0083"]
}

Reference Accumulation: As the camera moves through the scene, reference frames are accumulated to allow each target frame to know which reference images are relevant based on position and viewing angle.

---

Dataset Statistics

CARLA Stage2 Dataset Statistics (Pedestrian Only)

Town	Mode	Scenes	Images	Avg Images/Scene
Town01	pedestrian	480	96,000	200.0
Town01	TOTAL	480	96,000	200.0
Town02	pedestrian	488	97,600	200.0
Town02	TOTAL	488	97,600	200.0
Town03	pedestrian	480	96,000	200.0
Town03	TOTAL	480	96,000	200.0
Town04	pedestrian	490	98,000	200.0
Town04	TOTAL	490	98,000	200.0
Town05	pedestrian	480	96,000	200.0
Town05	TOTAL	480	96,000	200.0
Town06	pedestrian	490	98,000	200.0
Town06	TOTAL	490	98,000	200.0

Grand Total

Metric	Value
Total Scenes	2,908
Total Images	581,600
Towns	6 (Town01-06)

---

Example Usage

Installation

pip install datasets torch pillow numpy webdataset

Basic Usage with webdataset

import webdataset as wds
import json
import numpy as np
import io

# Stream dataset from HuggingFace
url = "https://huggingface.co/datasets/mkxdxd/carla-dataset-ped/resolve/main/Town01/pedestrian/{carla-stage2-000000..carla-stage2-000010}.tar"

dataset = wds.WebDataset(url).decode("pil")

for sample in dataset:
    key = sample["__key__"]
    rgb = sample["rgb.png"]           # PIL.Image (1280x704)
    depth = np.load(io.BytesIO(sample["depth.npy"]))  # numpy array (704x1280)
    camera = json.loads(sample["camera.json"])        # dict
    metadata = json.loads(sample["metadata.json"])    # dict
    
    print(f"Key: {key}")
    print(f"RGB size: {rgb.size}")
    print(f"Depth shape: {depth.shape}")
    print(f"Scene: {metadata['scene_id']}, Town: {metadata['town']}")
    break

Dataset generated from CARLA Simulator