demalenk/itto-dataset

Is This Tracker On? A Benchmark Protocol for Dynamic Tracking

🌐 Project Website | 💻 Code | 📚 Paper

Dataset card for ITTO

ITTO is a challenging new benchmark suite for evaluating and diagnosing the capabilities and limitations of point tracking methods.

Installation:

ITTO contains three component datasets: (1) MOSE, (2) L-VOS, (3) Ego4D. You will have to install the L-VOS and Ego4D components of the dataset individually due to licensing permissions.

1. First, install the dataset:

git clone https://huggingface.co/datasets/demalenk/itto

2. Next, install the L-VOS portion. Note that this will create temporary files in intermediate steps:

bash lvos/install_lvos.sh

3. Next, install the Ego4D portion of the dataset. Note that you have need access to obtain access to Ego4D data. License requests can take a few hours to a few days, and can be obtained here: https://ego4d-data.org/docs/start-here/

Make sure that you have the ego4d CLI installed by running:

pip install ego4d

bash ego4d/install_ego4d.sh

Running Evaluations with ITTO

We provide evaluation scripts in the ITTO github repo, which contains dataloaders and model evaluation scripts for the numbers reported in the paper.

Citing ITTO

ITTO aggregates videos from three public datasets: MOSE, LVOS, and Ego4D. Please cite all of them in addition to citing ITTO. Each component keeps its original license and usage terms. We provide all license information in LICENSE.md.