New standalone tool: standalone_icon_extractor.py
A focused, single-purpose GUI tool for extracting Entropia Universe icons:
- Auto-detects cache folder including version subfolders
- Converts TGA to PNG with transparent backgrounds
- Canvas sizing options (64x64 to 512x512)
- Upscale options: None, NEAREST, HQ4x, LANCZOS
- Selective conversion (select specific files or all)
- Saves settings between sessions
- Dark theme UI
Usage:
python standalone_icon_extractor.py
Does not require Lemontropia Suite to be installed/running.
1. Auto-detection now uses rglob to find TGA files in ALL subfolders
- Properly finds icons in version subfolders like 19.3.2.201024
- Shows relative paths in the file list
2. Default settings changed:
- Default canvas: 320x320 (index 5)
- Default upscale: No Upscaling (first in list)
3. Added iconcache.dat parser module (icon_cache_parser.py)
- Attempts to extract item names from binary cache file
- Uses heuristics to find icon ID to name mappings
- New option: "❌ No Upscaling" - keeps original icon size, just centers on canvas
- Useful when you want the icon at original resolution on a larger canvas
- Tooltip updated to explain all options
- When "none" is selected, upscale parameter becomes False
- New module: modules/ai_upscaler.py
- Real-ESRGAN integration for AI-powered upscaling
- Designed specifically for low-res game textures/icons
- 4x upscale factor with neural network enhancement
- Falls back gracefully if model not available
- Updated TGA Converter UI:
- New "AI Enhanced (Real-ESRGAN)" option in upscale dropdown
- Only shown if Real-ESRGAN is installed
- AI upscaling happens before canvas placement
- Shows progress during AI model loading
Real-ESRGAN is much better than basic methods for rendered (non-pixel) icons!
To use AI upscaling:
1. pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
2. pip install realesrgan
3. Download RealESRGAN_x4plus.pth model
4. Select "AI Enhanced" in the UI
- NEAREST: Sharp pixel edges (best for clean pixel art)
- HQ4x-style: Integer scale first, then smooth to target with edge enhancement
- LANCZOS: Very smooth (best for photorealistic textures)
- Added upscale method selector to UI
- Each method uses UnsharpMask for enhanced crispness
- Use integer scaling (2x, 3x, 4x) instead of fractional for sharp pixels
- Use NEAREST neighbor resampling (better for pixel art than Lanczos)
- Add sharpening filter after upscale
- Reduced padding to 85% so icons fill more of canvas
This produces much sharper results for game icons!
- Now checks 2 levels deep for TGA files (handles version subfolders like 19.3.2.201024)
- convert_cache_folder now uses rglob to find TGA files in all subfolders
- Will auto-detect the version folder in ProgramData
The _apply_canvas method existed but was never called in convert_tga_to_png.
Now properly applies canvas sizing in both PIL and manual conversion paths.
- Removed WebP support (PNG only now)
- Added canvas size selector: Original, 64x64, 128x128, 256x256, 280x280, 320x320, 512x512
- Icons are centered on transparent canvas
- Added 'Upscale small icons' checkbox to scale up small icons to better fill canvas
- Uses high-quality Lanczos resampling for upscaling
- Default canvas size: 320x320
- Updated TGAConverter.convert_tga_to_png() to support 'webp' format
- High quality WebP encoding (quality=95, method=6)
- Added format selector in UI (PNG or WebP)
- Convert button text updates based on selected format
- WebP is better for web uploads: smaller files, better quality
WebP advantages:
- 25-35% smaller file size than PNG
- Better quality at same file size
- Alpha channel support
- Widely supported in modern browsers
- PIL handles more TGA formats correctly
- Manual parser fixed to properly skip ID and color map fields
- Added better error messages with expected vs actual data size
- Fallback chain: PIL first, then manual parser
- Worker now accepts cache_path parameter
- _start_conversion passes the selected path to worker
- Fixes 'Cache folder not found' error during conversion
- Added: C:\ProgramData\Entropia Universe\public_users_data\cache\icon
- Now searches version subfolders (e.g., 19.3.2.201024)
- Also searches ProgramData recursively as fallback
- New module: modules/tga_converter.py
- Reads Truevision TGA format from game cache
- Converts to PNG with proper color handling (BGR->RGB)
- Auto-detects cache folder location
- Batch conversion support
- Handles alpha channels
- New dialog: ui/tga_converter_dialog.py
- Visual interface for TGA conversion
- Shows list of cached .tga files
- Progress tracking during conversion
- One-click convert all
- Added to Tools → Computer Vision → TGA Icon Converter (Ctrl+T)
This is much more efficient than extracting icons from screenshots!
The game already has the icons in cache as .tga files.
Changed default locations:
- Screenshots: Documents/Entropia Universe/Screenshots
- Icons: Documents/Entropia Universe/Icons
This provides better organization and easier access for users.
All directories are created automatically if they don't exist.
- Add Directories group to General tab
- Settings for screenshots directory
- Settings for extracted icons directory
- Browse buttons for both directories
- Values saved to QSettings
- New module: modules/inventory_scanner.py
- Detects inventory and item details windows
- Extracts item icons from inventory grid
- Reads item stats from details panel
- Parses weapon, armor, and common stats
- Handles scrolling (planned for future)
- New dialog: ui/inventory_scanner_dialog.py
- Visual interface for scanning
- Shows extracted icons in grid view
- Displays item stats in table
- Saves results to JSON
- Background worker for non-blocking scans
- Updated main_window.py:
- Added Tools → Computer Vision → Inventory Scanner (Ctrl+I)
- Integrated with existing GameVisionAI
This allows users to extract item data from Entropia Universe
for gear management and loadout configuration.
- Create new ui/settings_dialog.py with tabbed interface
- Tabs: General, Screenshot Hotkeys, Computer Vision, Advanced
- Remove old SettingsDialog from main_window.py
- Update on_settings() to use new comprehensive dialog
- Screenshot hotkey settings now in Settings (Ctrl+,)
- Computer Vision settings now in Settings dialog
- Remove persistent mss.mss() instance from AutoScreenshot
- Use context manager 'with mss.mss() as sct:' for each capture
- Fixes '_thread._local' object has no attribute 'srcdc' error
- Allows screenshots to work from hotkey callbacks in background threads
SQLite's TIMESTAMP type causes auto-conversion issues when data is in ISO format
- Changed all TIMESTAMP columns to TEXT in schema.sql
- Disabled detect_types in database.py to prevent auto-conversion
- This stores datetime as ISO strings without parsing issues
- Convert sqlite3.Row to list before accessing by index
- Add detailed debug logging for troubleshooting
- Fix fromisoformat error by checking for None values
- Add try/except around each row to skip bad data
- Changed from dict-style access (proj['type']) to index access (proj[0])
- Added try/except around each row processing to skip bad rows
- Added detailed error logging with traceback
- Handle cases where row has fewer columns than expected
- Add '-- Select a template --' as first item in combo box
- Block signals while refreshing to prevent unwanted callbacks
- Select newly created template automatically
- Fix index check to skip placeholder (index > 0 instead of >= 0)
- Add enable/disable based on whether templates exist
- Rewrote screenshot_hotkey.py to use Qt QShortcut as primary method
- Qt shortcuts work reliably when app is focused (no admin needed)
- Added optional 'keyboard' library for global hotkeys (requires admin)
- Updated main_window.py to pass parent window to hotkey manager
- Shows clear status: global vs Qt-only mode
- Users can now press F12 etc when app is focused
- Use pynput format <f12>, <shift>+<f12> etc for hotkey parsing
- Initialize screenshot hotkeys AFTER UI setup (fixes log_output error)
- Clean up hotkey display in help text (remove angle brackets)
- Add modules/screenshot_hotkey.py with global hotkey listener using pynput
- Default hotkeys: F12=Full, Shift+F12=Region, Ctrl+F12=Loot, Alt+F12=HUD
- Integrate into main_window.py with Settings dialog
- Hotkeys work globally (even when game is focused)
- Add Tools → Screenshot Hotkeys menu (Ctrl+Shift+S)
- Shows status message when screenshot is captured
Based on PyImageSearch article: https://pyimagesearch.com/2022/03/14/improving-text-detection-speed-with-opencv-and-gpus/
- Add opencv_text_detector.py using EAST model with CUDA support
- OCRProcessor now uses PaddleOCR first, falls back to OpenCV
- OpenCV text detection: ~97 FPS on GPU vs ~23 FPS on CPU (4x faster)
- No heavy dependencies (PyTorch/PaddlePaddle) required for OpenCV mode
- Auto-downloads EAST model on first use
- Add early torch import before PyQt6 imports in main_window.py
- This prevents the c10.dll initialization error on Windows
- Solution 3 from user research: import torch before PyQt
- Add fix_pytorch.bat for Windows users to repair PyTorch installation
- Add verify_vision.py to check and auto-fix vision dependencies
- Make torch import optional in game_vision_ai.py to prevent crashes
- Provides graceful fallback if PyTorch fails to load
- Code was calling db.fetchall() and db.fetchone() but these methods didn't exist
- Add helper methods that execute query and return results
- Fixes template loading and other database operations
- Fix recursion error in GalleryDialog by blocking signals during session filter updates
- Fix missing 'avatar_name' field in SetupWizard by registering the field properly
- Both dialogs should now work without errors
- Check evade type for 'target Evaded' or 'target Dodged'
- Count weapon cost on those events (you fired but missed)
- Armor decay still only counted on actual damage taken
- Fixes: weapon cost now accurate for all shots including misses
- Changed from 0.0005 to 0.00025 PED per damage (0.025 PEC)
- Based on user feedback: 504 damage should give ~0.14 PED decay
- Removed complex fallback logic, using simple accurate formula
- Formula: damage * 0.00025 = armor decay in PED
LemonNexus