# Temporal Logic Video (TLV) Dataset
## Overview
## Overview
The Temporal Logic Video (TLV) Dataset addresses the scarcity of state-of-the-art video datasets for long-horizon, temporally extended activity and object detection. It comprises two main components:
1. Synthetic datasets: Generated by concatenating static images from established computer vision datasets (COCO and ImageNet), allowing for the introduction of a wide range of Temporal Logic (TL) specifications.
2. Real-world datasets: Based on open-source autonomous vehicle (AV) driving datasets, specifically NuScenes and Waymo.
## Table of Contents
- [Dataset Composition](#dataset-composition)
- [Dataset (Release)](#dataset)
- [Installation](#installation)
- [Usage](#usage)
- [Data Generation](#data-generation)
- [Contribution Guidelines](#contribution-guidelines)
- [License](#license)
- [Acknowledgments](#acknowledgments)
## Dataset Composition
### Synthetic Datasets
- Source: COCO and ImageNet
- Purpose: Introduce artificial Temporal Logic specifications
- Generation Method: Image stitching from static datasets
### Real-world Datasets
- Sources: NuScenes and Waymo
- Purpose: Provide real-world autonomous vehicle scenarios
- Annotation: Temporal Logic specifications added to existing data
## Dataset
Though we provide a source code to generate datasets from different types of data sources, we release a dataset v1 as a proof of concept.
### Dataset Structure
We provide a v1 dataset as a proof of concept. The data is offered as serialized objects, each containing a set of frames with annotations.
#### File Naming Convention
`\:source:\-number_of_frames:\-\.pkl`
#### Object Attributes
Each serialized object contains the following attributes:
- `ground_truth`: Boolean indicating whether the dataset contains ground truth labels
- `ltl_formula`: Temporal logic formula applied to the dataset
- `proposition`: A set of proposition for ltl_formula
- `number_of_frame`: Total number of frames in the dataset
- `frames_of_interest`: Frames of interest which satisfy the ltl_formula
- `labels_of_frames`: Labels for each frame
- `images_of_frames`: Image data for each frame
You can download a dataset from here. The structure of dataset is as follows: serializer
```
tlv-dataset-v1/
├── tlv_real_dataset/
├──── prop1Uprop2/
├──── (prop1&prop2)Uprop3/
├── tlv_synthetic_dataset/
├──── Fprop1/
├──── Gprop1/
├──── prop1&prop2/
├──── prop1Uprop2/
└──── (prop1&prop2)Uprop3/
```
#### Dataset Statistics
1. Total Number of Frames
| Ground Truth TL Specifications | Synthetic TLV Dataset | | Real TLV Dataset | |
| --- | ---: | ---: | ---: | ---: |
| | COCO | ImageNet | Waymo | Nuscenes |
| Eventually Event A | - | 15,750 | - | - |
| Always Event A | - | 15,750 | - | - |
| Event A And Event B | 31,500 | - | - | - |
| Event A Until Event B | 15,750 | 15,750 | 8,736 | 19,808 |
| (Event A And Event B) Until Event C | 5,789 | - | 7,459 | 7,459 |
2. Total Number of datasets
| Ground Truth TL Specifications | Synthetic TLV Dataset | | Real TLV Dataset | |
| --- | ---: | ---: | ---: | ---: |
| | COCO | ImageNet | Waymo | Nuscenes |
| Eventually Event A | - | 60 | - | - |
| Always Event A | - | 60 | - | - |
| Event A And Event B | 120 | - | - | - |
| Event A Until Event B | 60| 60 | 45| 494 |
| (Event A And Event B) Until Event C | 97 | - | 30 | 186|
## Installation
```bash
python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip build
python -m pip install --editable ."[dev, test]"
```
### Prerequisites
1. ImageNet (ILSVRC 2017):
```
ILSVRC/
├── Annotations/
├── Data/
├── ImageSets/
└── LOC_synset_mapping.txt
```
2. COCO (2017):
```
COCO/
└── 2017/
├── annotations/
├── train2017/
└── val2017/
```
## Usage
Detailed usage instructions for data loading and processing.
### Data Loader Configuration
- `data_root_dir`: Root directory of the dataset
- `mapping_to`: Label mapping scheme (default: "coco")
- `save_dir`: Output directory for processed data
### Synthetic Data Generator Configuration
- `initial_number_of_frame`: Starting frame count per video
- `max_number_frame`: Maximum frame count per video
- `number_video_per_set_of_frame`: Videos to generate per frame set
- `increase_rate`: Frame count increment rate
- `ltl_logic`: Temporal Logic specification (e.g., "F prop1", "G prop1")
- `save_images`: Boolean flag for saving individual frames
## Data Generation
### COCO Synthetic Data Generation
```bash
python3 run_scripts/run_synthetic_tlv_coco.py --data_root_dir "../COCO/2017" --save_dir ""
```
### ImageNet Synthetic Data Generation
```bash
python3 run_synthetic_tlv_imagenet.py --data_root_dir "../ILSVRC" --save_dir ""
```
Note: ImageNet generator does not support '&' LTL logic formulae.
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
## Citation
If you find this repo useful, please cite our paper:
```bibtex
@inproceedings{Choi_2024_ECCV,
author={Choi, Minkyu and Goel Harsh and Omama, Mohammad and Yang, Yunhao and Shah, Sahil and Chinchali and Sandeep},
title={Towards Neuro-Symbolic Video Understanding},
booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
month={September},
year={2024}
}
```
[contributors-shield]: https://img.shields.io/github/contributors/UTAustin-SwarmLab/temporal-logic-video-dataset.svg?style=for-the-badge
[contributors-url]: https://github.com/UTAustin-SwarmLab/temporal-logic-video-dataset/graphs/contributors
[forks-shield]: https://img.shields.io/github/forks/UTAustin-SwarmLab/temporal-logic-video-dataset.svg?style=for-the-badge
[forks-url]: https://github.com/UTAustin-SwarmLab/temporal-logic-video-dataset/network/members
[stars-shield]: https://img.shields.io/github/stars/UTAustin-SwarmLab/temporal-logic-video-dataset.svg?style=for-the-badge
[stars-url]: https://github.com/UTAustin-SwarmLab/temporal-logic-video-dataset/stargazers
[license-shield]: https://img.shields.io/github/license/UTAustin-SwarmLab/temporal-logic-video-dataset.svg?style=for-the-badge
[license-url]: https://github.com/UTAustin-SwarmLab/temporal-logic-video-dataset/blob/master/LICENSE.txt
