Data & Code

Repos & Datasets

Every featured paper ships with a repo and, where licensable, a dataset. Stars and last-commit dates are cached at build time. If a paper is missing, the code is private or pending release — open an issue and we'll surface it.

Repositories

All public repos →

RepositoryDescriptionLang★Updated

benchoi93/TrajGAIL Urban vehicle trajectory generation via generative adversarial imitation learning (TR-C 2021). Python ★ 68 Sep 2021 UMN-Choi-Lab/BO4Mob Bayesian optimization benchmarks for high-dimensional urban mobility (NeurIPS 2025 Datasets & Benchmarks). Python ★ 18 Apr 2026 UMN-Choi-Lab/DGMinTransportation Companion code & notebooks for the TR-C 2025 deep generative models survey. Jupyter ★ 9 Jul 2025 UMN-Choi-Lab/TrajFlow Normalizing-flow framework for occupancy density estimation from trajectories (TrajFlow). Python ★ 3 May 2026 UMN-Choi-Lab/TrafficNetQA QA benchmark for evaluating LLM performance on traffic network files (SIGSPATIAL 2025). Python ★ 1 Nov 2025 UMN-Choi-Lab/camera2detector Real-time traffic-camera validation of MnDOT highway detectors using YOLO computer vision. Python ★ 1 Apr 2026 UMN-Choi-Lab/HighwayVLM Open-source highway-safety analysis with vision-language models (UROP, Spring 2026). Python ★ 1 Apr 2026 UMN-Choi-Lab/ODS_PLD Analytical OD estimation via NNLS and Projected Langevin Dynamics, validated under SUMO on the BO4Mob benchmark. Python ★ 0 Apr 2026 UMN-Choi-Lab/SynPopPred LLM-based synthetic population generation — tutorial comparing IPF, CTGAN, DistilGPT-2, and Llama 3.1 8B. Python ★ 0 Mar 2026

Canonical list at github.com/UMN-Choi-Lab ↗. ★ counts & dates cached 2026-06-03 · refreshed periodically, not via live API.

Datasets

MTOdata

Curated mobility & transportation operations datasets, hosted at mtodata.github.io.

license · varies cite · per-dataset Open ↗

PeMS-BAY-2022

Updated PeMS-BAY traffic-speed dataset (Bay Area, 2022) — a refresh of the widely-used PeMS-BAY benchmark for spatiotemporal forecasting.

license · open cite · UMN Choi Lab Open ↗

Reproducibility

We commit to the following for every released repo:

Pinned dependencies · requirements.txt or environment.yml
Single-command training entrypoint · scripts/train.sh
Single-command evaluation · scripts/eval.sh
Pre-trained checkpoints when license permits
Dataset DOI or download script
Hardware & wall-clock from the paper (partial)
Hyperparameter search logs (partial)

Issues for missing artifacts: file at github.com/UMN-Choi-Lab/<repo>/issues — tag with repro.