Getting Started
Participants register a team, receive a unique team ID, download the data and Starter Kit, generate their model predictions, and submit to the benchmark.
Step-by-Step Participation
-
Register Team
Register or manage your team to receive a unique team ID, updates, and access to submission instructions.
Register or manage team -
Receive Unique Team ID
After registration, each team is issued a unique identifier that must be used for all submissions and leaderboard entries.
-
Download Data
Download the Spring and RobustSpring training splits needed for model development, local validation, and practice runs.
-
Install the Starter Kit
Set up the PTLFlow-based Starter Kit with training and inference scripts, data loaders, and example configurations.
-
Run Baseline
Run the provided baseline configurations to verify your environment, data paths, and end-to-end inference pipeline.
-
Generate Predictions
Produce model outputs for your chosen task(s) and export them in the required prediction format.
-
Validate Archive Locally
Package your predictions into a submission archive and run local validation to catch formatting or naming issues before uploading.
-
Submit to Benchmark Server
Upload your validated archive to the benchmark to receive scores on the development leaderboard.
-
Submit Workshop Paper and Reproducibility Package
Prepare and submit a 4–6 page workshop paper describing your method and experiments. With the camera-ready deadline, include a reproducibility package with dataset declarations, training and evaluation scripts, resource requirements, and a pinned environment setup. See evaluation requirements for details.
PTLFlow-Based Tools and Scripts
RoCo-Spring Starter Kit repository: hmorimitsu/roco-spring-devkit
Clone the official RoCo-Spring Starter Kit for its PTLFlow-based tools and scripts. Challenge submission requirements and validation steps are maintained on this website.
If you use the PTLFlow-based Starter Kit or PTLFlow components in your work, please cite the following reference.
@misc{morimitsu2021ptlflow,
author = {Henrique Morimitsu},
title = {PyTorch Lightning Optical Flow},
year = {2021},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/hmorimitsu/ptlflow}}
}Getting Help
GitHub issues: Report a website or challenge issue
Questions and discussion: Start a public support thread