As of 6 August 2026 this project is closed. There will be no further releases, bug fixes, or support, and issues and pull requests are no longer being handled.
The source stays online and remains available under its existing licence, so anyone is welcome to fork it and continue the work.
The screenshot-recognition half lives on as a standalone project: sto-warp.
Final release: v3.0 (17 May 2026).
A build planning and screenshot recognition tool for Star Trek Online.
SETS-WARP combines three tools in one package:
SETS (STO Equipment and Trait Selector) — build planner for Star Trek Online. Plan ship and ground loadouts, skill trees, and bridge officer assignments without owning the items in-game. Builds can be shared as PNG and JSON files.
All credits for SETS to Shinga — STOCD/SETS.
WARP (Weaponry & Armament Recognition Platform) — reads your in-game screenshots and automatically fills in your SETS build. Detects equipment, traits, bridge officers, and ship information using computer vision and machine learning.
WARP CORE — trainer interface built into WARP. Review and correct recognition results, confirm annotations, and retrain the local ML models (Icon Classifier + Layout Regressor) on your own data to improve accuracy over time.
Autonomous Recognition Engine (P0-P5) — WARP now features an intelligent inference pipeline that learns your window layout and ship configurations. It includes smart OCR for ship metadata (name/type/tier), slot position inference, multi-config layout memory, dynamic anchor recalibration, and a CNN-based layout regression model.
→ See WARP_GUIDE.md for a full usage guide with step-by-step instructions for WARP and WARP CORE.
→ See SETS_GUIDE.md for SETS build planner documentation.
Download sets-warp-vX.Y-setup.exe from the Releases page and run it.
- Installs to
%LOCALAPPDATA%\SETS-WARP— no administrator rights required - Creates Start Menu and optional Desktop shortcut
- On first launch, chooses between SETS only (~3 GB) or SETS + WARP (~10 GB) and downloads all dependencies automatically
- Post-install wizard warns if Python 3.11+ is missing from PATH
git clone https://github.com/raman78/sets-warp.git
cd sets-warp
sets_warp.batgit clone https://github.com/raman78/sets-warp.git
cd sets-warp
./sets_warp.shOn first run a graphical setup window appears asking you to choose:
| Option | Disk space | What's included |
|---|---|---|
| SETS only | ~3 GB | Build planner, ship/equipment database |
| SETS + WARP | ~10 GB | Everything above + screenshot recognition, ML training |
The installer then downloads a portable Python runtime and all required packages automatically. Subsequent launches complete in under a second.
Desktop integration is set up automatically on first run — SETS-WARP appears in your application launcher with its own icon.
tkinter requirement (first run only): tkinter is needed to display the setup window. If it's missing, the script prints the exact install command for your distro (apt / dnf / pacman / zypper / apk).
git clone https://github.com/raman78/sets-warp.git
cd sets-warp
./sets_warp.shSame automatic setup as Linux. No desktop integration step on macOS.
The setup window walks through these steps automatically:
- Download portable Python 3.13 (~65 MB, one-time)
- Create isolated virtual environment (
.venv/) - Install Python packages from PyPI (~3 GB SETS-only / ~10 GB SETS+WARP, one-time)
- (SETS+WARP only) Build WARP item database from STO cargo data
- (SETS+WARP only) Download community ML model — so WARP works immediately without manual training
The choice between SETS-only and SETS+WARP is saved and can be changed at any time:
- Settings → Installation → SETS + WARP checkbox
- Check (switch to SETS + WARP): on the next launch,
bootstrap.pydownloads and installs ~10 GB of ML dependencies (PyTorch, EasyOCR, OpenCV, HuggingFace libs). - Uncheck (switch to SETS-only): on the next launch, WARP packages (~7 GB) are automatically removed from the virtual environment.
- A confirmation dialog appears before the change is committed. The switch takes effect after the app restarts.
- Check (switch to SETS + WARP): on the next launch,
git pull
./sets_warp.sh # Linux / macOS
sets_warp.bat # Windows (git install)SETS-WARP checks GitHub Releases for a new version 3 seconds after launch (background thread, non-blocking). When a newer release is found, an update dialog appears:
- Update now — pulls the update and restarts automatically:
- Git install (
git clone): runsgit pullthen restarts - Zip / installer install: downloads the release zip from GitHub, extracts it in-place, then restarts
- Git install (
- Later — dismisses the dialog; the check runs again on the next launch
- Don't remind me for vX.Y — snoozes notifications for that specific release; the check still runs, but that version is silently skipped
To disable automatic update checks entirely, go to Settings → SETS-WARP Updates and uncheck Check for updates automatically. The installed version is shown there as well. There is no manual "check now" button — to force an immediate check, restart the app.
bootstrap.py detects and installs new or updated Python dependencies on every launch — no manual pip install needed after git pull.
The community-trained icon recognition model is updated separately from the app. 15 seconds after launch, a background check contacts the backend to compare model timestamps. If the remote model is newer than the locally installed one, it is downloaded silently from HuggingFace (sets-sto/warp-knowledge) and loaded automatically — no restart needed.
- Rate-limited to at most once per 24 hours
- A locally retrained model (via WARP CORE → Train Model) always takes priority — if you trained more recently than the remote, no download occurs
screen_classifier.ptis also downloaded automatically if it is missing
- Take screenshots of your in-game build screens (equipment, traits, bridge officers — separately or as one combined screenshot). One build per folder.
- Click the ⚡ WARP button, select the screenshot folder.
- WARP detects your ship, reads slot counts, recognises each item, and fills your build automatically.
- Use WARP CORE to review results, correct mistakes, and confirm items — building up training data that improves recognition over time.
→ Full instructions: WARP_GUIDE.md → ML pipeline details: ML_PIPELINE.md
Local model: Confirmed items in WARP CORE are saved as training data. After confirming enough items, click Train Model to retrain the icon classifier on your corrections.
Community model: Confirmed items are sent anonymously to a shared knowledge base. The central model is retrained hourly from community contributions and downloaded at the next startup. Fresh installs get the latest community model automatically during setup.
→ Full technical details: ML_PIPELINE.md
Go to Settings → Uninstall → Uninstall SETS-WARP. After confirmation the app exits and a cleanup script runs in the background that:
- removes the installation folder (including
.venv/and all downloaded data) - removes the desktop entry (Linux)
This is permanent and cannot be undone. On Windows (installer), use the standard Add/Remove Programs entry instead.
All credits for SETS to Shinga — STOCD/SETS.
Item data from the official STO wiki. If something is missing or wrong, update it there.
Bug reports and suggestions: GitHub Issues or STOBuilds Discord → #sets-support.
| File | Contents |
|---|---|
| WARP_GUIDE.md | WARP and WARP CORE user guide |
| SETS_GUIDE.md | SETS build planner guide |
| ML_PIPELINE.md | ML pipeline technical reference |
| CHANGELOG.md | Release history |