A build planning and screenshot recognition tool for Star Trek Online.
Status: v2.0b — tested on Linux and Windows. macOS support is present but untested.
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 |