Skip to content

Roadmap

This roadmap tracks high-priority milestones to mature the Awesome Human Activity Recognition list into a best-in-class open resource. Timelines are quarter-based targets; adapt as contributors join.

v0.2.0 — Q1 2026 ✅ (Current Release)

  • [x] Publish 40+ dataset cards spanning all modalities with canonical references.
  • [x] Add MIT LICENSE file and fix all badge/citation URLs.
  • [x] Finalize link validation pipeline (GitHub Action using lycheeverse/lychee).
  • [x] Add Markdown lint CI workflow for consistent formatting.
  • [x] Fix data inconsistencies across README, dataset cards, and i18n files.
  • [x] Update roadmap to reflect actual project state.

v0.3.0 — Q2 2026

  • Launch GitHub Discussions with channels for dataset requests and SOTA tracking.
  • Ship tools/catalog_builder.py JSON + CSV export with GitHub Action auto-generation.
  • Release quick-start Jupyter notebooks:
  • Wearable HAR baseline on PAMAP2 / UCI-HAR (PyTorch).
  • Video action recognition on Kinetics-700 (PyTorchVideo).
  • Introduce automated badge generation (dataset counts, modalities) via GitHub Action + Shields endpoint.
  • Provide standardized dataset schema (YAML frontmatter) with validation to support API consumers.

v0.4.0 — Q3 2026

  • Add baseline reproducibility scripts (PyTorch) covering at least:
  • Skeleton-based GCN on NTU RGB+D 120.
  • Transformer video model on Something-Something V2.
  • Self-supervised wearable baseline on WISDM.
  • Expand dataset coverage to synthetic human simulations (e.g., SAPIEN, Habitat, ManiSkill2).
  • Introduce monthly digest (via Discussions) summarizing new datasets, papers, and PR highlights.
  • Provide standardized dataset schema (YAML/JSON) with validation to support API consumers.

v1.0.0 — Q4 2026

  • Publish MkDocs documentation site served via GitHub Pages, synced with repository metadata.
  • Build interactive web explorer with filtering by modality, license, tasks, and download size.
  • Offer dataset health metrics (link uptime, missing files, version notes).
  • Host community benchmarking challenge (e.g., wearable activity recognition transfer) with partner labs.

Beyond

  • Collaborate with dataset authors to host official mirrors or metadata updates.
  • Explore plugin integrations (Weights & Biases, Hugging Face Datasets) to keep entries live.
  • Apply to awesome list for official inclusion.
  • Multilingual dataset card translations (beyond README-level i18n).

Contributors can propose roadmap updates via issues tagged type:roadmap. Priorities adjust as the community scales.