MASIV Multi-Sequence Dataset

Toward Material-Agnostic System Identification from Videos

Yizhou Zhao¹, Haoyu Chen¹, Chunjiang Liu¹, Zhenyang Li², Charles Herrmann³, Junhwa Hur³, Yinxiao Li³, Ming‑Hsuan Yang⁴, Bhiksha Raj¹, Min Xu^1*

¹Carnegie Mellon University ²University of Alabama at Birmingham ³Google ⁴UC Merced

Introduction

The MASIV Multi-Sequence Dataset is a synthetic dataset generated by Genesis to evaluate the generalization of data-driven constitutive models. The dataset contains 10 objects of 5 distinct materials (Elastic, Elastoplastic, Liquid, Sand, and Snow), with each object including 10 different multi-view sequences. These sequences have randomized initial conditions for location, pose, and velocity. Each sequence includes 11 views, with each view consisting of 30 frames.

Dataset Structure

MASIV/
├── 0_0/                                            # <ObjectID>_<SequenceID>
  ├── data/                                         # Contains per-frame images
  ├── point_clouds/                                 # Stores point cloud data for 30 frames
  ├── videos/                                       # Stores 11 videos, one for each view
  ├── all_data.json                                 # A JSON file containing camera information
  ├── metadata.json                                 # A JSON file containing global simulation and object-specific parameters
  ├── transforms_test.json                          # Contains the camera transformation matrices and file paths for the images in the test set
  ├── transforms_train.json                         # Contains the camera transformation matrices and file paths for the images in the training set
  ├── transforms_val.json                           # Contains the camera transformation matrices and file paths for the images in the validation set
├── 0_1/
...
├── 9_9/ 

Citing MASIV

If you find this dataset useful in your work, please consider citing our paper:

@article{zhao2025masiv,
  title={MASIV: Toward Material-Agnostic System Identification from Videos},
  author={Zhao, Yizhou and Chen, Haoyu and Liu, Chunjiang and Li, Zhenyang and Herrmann, Charles and Hur, Junhwa and Li, Yinxiao and Yang, Ming-Hsuan and Raj, Bhiksha and Xu, Min},
  journal={arXiv preprint arXiv:2508.01112},
  year={2025}
}