Datasets
Download ZIP — 387 MB
apple
Preview
100 views · 800×800 · PLY · 387 MB
Dataset Specifications
| Views | 100 (Full sphere, Fibonacci sampling) |
| Resolution | 800 × 800 px (transparent) |
| Modalities | RGB Depth 8-bit Depth 16-bit Normals Masks |
| Point Cloud | points3D.ply |
| Format | nerfstudio (transforms.json) |
| License | — |
| Download Size | 387 MB |
Download Dataset
387 MB
Need higher resolution or more views? Render your own datasets in the portal with configurable tiers up to 400 views at 2048×2048.
Contents
dataset.zip
├── images/ — 100 RGB views (PNG, transparent)
├── depth/ — 8-bit grayscale depth maps
├── depth_16bit/ — 16-bit grayscale depth maps
├── normals/ — Surface normal maps
├── masks/ — Binary object masks
├── points3D.ply — Point cloud
├── transforms.json — Camera poses (nerfstudio format)
└── overview.webp — Contact sheet preview
Quick Start
# Train with nerfstudio
ns-train nerfacto --data path/to/dataset
# Load camera poses
import json
with open("transforms.json") as f:
meta = json.load(f)
print(f"{len(meta['frames'])} views at {meta['w']}x{meta['h']}") Citation
@misc{dxgl_apple_2026,
title = {apple — Multi-view Dataset},
author = {DXGL},
year = {2026},
url = {https://dx.gl/datasets/vCHDLxjWG65d},
note = {100 views, 800x800}
}