Dlta-ai

This release includes new features such as a user guide, video frame extraction, and full custom export support. It also includes bug fixes and UI enhancements.

New Features 🌟
- User Guide is now available in the tool, you can access it from the Help menu or from [here](https://0ssamaak0.github.io/DLTA-AI/)
- New Input Option: Extracting Video to Frame, open video as frames in directory mode, you can select start and end frames as well as sampling rate (every 2 frames, every 3 frames, etc.)
- Custom Export is now fully supported in all modes, `custom_exports.py` is now updated with detailed documentation and examples
- Adding splash screen while loading
- New icons for Edit and Context Menu (previously was using the same icon for all funcionalities)
- New UI for Edit Labels and Delete Polygons menus
- General UI Enhancements, Buttons name changes, ToolTips are more clear.
- Adding slider to Confidence and IOU Thresholds
- Adding Visualization Options to Image and Dir modes
- Adding Confidence visualization option to all modes
- New Tracking Interpolation UI allows Both Interpolation Methods (Linear, SAM) can be done between (Detected frames, keyframes) and option to overwrite used frames
- Shortcuts menu now has `reset` and `restore defaults` buttons
- Adding `Add Custom Export` button to export dialog opens `custom_exports.py` in your default editor
- Adding option to Remove Keyframes

Bug Fixes 🐞
- Tracking after closing the tool now starts from last tracked ID (previously was starting from 1)
- removing uncenessary prints in SAM, logs are more clean now
- Enhance Polygons in SAM toolbar and context menu are unified now (same name, icon and shortcut)
- Fixing a bug when changing the confidence of a custom class polygon
- Fixing a bug shows a notification (Downloaded Successfully) even if the download failed
- Fixing a bug in creating first SAM instance
- Fixing a bug when deleting a non completed polygon with `ESC` key
- Fixing a bug in export adds COCO classes as custom classes (class_id > 80)
- Export in image mode no longer exports all other JSONs in the same directory
- Removing `WindowContextHelpButtonHint` from all dialogs, since it's inconvenient and has no functionality
- File List Dock is cleared when opening image or video or closing the current Image in directory mode, it's opened shifted to the right by default to focus on the image names not the path
- Fixing a bug allows user to input 0 threshold
- Opening any new input clears Label list from old labels
- Visualization Font size is now a function of the image size, no more accidental very small or very big font sizes
- Loading Font size is also a function of the image size
- Fixing a bug in visualization in dir mode doesn't show polygons and other visualization options
- Changing Hard Coded Shortcuts
- Fixing a bug crashes the tool when clicking and releasing a selected shape very fast (very rare bug)
- SAM Interpolation now doesn't edit already made polygons when interpolating between frames
- Fixing a bug visualization options to be more responsive after edits
- Fixing a bug in export Error dialog
- If Export as Video is interrupted, the video is created with the frames that were already exported
- Export as video now show "Processing..." For the frames not being exported (previously was showing last exported frame number)
- Exporting an Empty video will result in an error dialog
- Custom Classes now have different colors in visualization

New Features 🌟
- Automatic **Check for Updates** option is add to help menu instead of **Version**


Bug Fixes 🐞
- yolon is installed by default to prevent error when running the app for the first time

New Features 🌟
- DLTA-AI can now be installed using pip using the command`pip install DLTA-AI`
- Enhance With SAM now works on all shapes if no shapes are selected
- delete is added again to right click and edit menus
- RAM info is added to Runtime data
- internal code documentation is added for (export, custom_export, model explorer, select classes, shortcut dialog, user preferences)
- notifications are enabled by default

Bug Fixes 🐞
- fixing normal interpolation error
- fixing set as key frame error (in interpolation)
- fixing error in scaling of shapes
- MOT export now reads and writes in the same time, no memroy overflow can happen
- version is now shown in the about dialog correctly (previously static)
- to address RAM overflow, memory usage info is printed to the console in tracking

Segment Anything (SAM) 🪄
- Adding Meta's new model: **Segment Anything**
- Support Point mode (Add / Remove)
- Support Box Mode
- Works in both images and videos
- SAM models can be downloaded in the Model Explorer
- Shortcuts for Controlling SAM
- SAM is supported in tracking (interploate with SAM) and works with many objects at once
- SAM can be used to enhance the segmentation of other models, even manual segmentations

Other Features 🌟
- Tracking now works with SAM, you can interpolate between frames with SAM
- SAM interpolation works in batch mode i.e, you can interpolate multiple objects at once
- saved_models.json now dynamically generated from downloaded models
- Allow user to Export to custom path instead of the original data path
- Adding Area to the exported COCO format (supports both bbox and segmentation)
- Optimization to Merge Models Feature (Faster)
- Adding a script to evaluate the exported COCO to Ground Truth using COCO API (outside the tool)
- Adding Non Maximum Suppression (NMS) to YOLOv8 models
- set threshold shows previous value by default for both Confidence and IOU
- adding option to Stop Tracking by pressing `Esc` key or `Stop Tracking` button or Closing the Window
- Merge Models now supports all mode (image, dir, video)
- Using ORJSON insteadd of nomral JSON to make read and write incredibly faster
- Menus are now more organized, new icons and shortcuts added
- faster video navigation
- more detaild export Dialog
- Export now supports custom classes (not in COCO)
- Optimization to interpolation
- allow user to add custom export formats which can be added in `utls.custom_exports.py`
- opening a previously annotated video enables export by default
- Select classes is now 3-column layout instead of one
- Select classes has Select / Deselect buttons
- Select classes has option to save default classes for future use
- Video toolbar is now less cluttered and more intuitive to use with input box and dropdown menu
- New Dock added for Video Visualization options
- support notifications for long tasks (if the app isn't focused)
- Preferences settings added (theme / mute notifications)
- shortcut settings added

Bug Fixes 🐞
- Fixing a bug in Directory mode (YOLOv8 models)
- Fixing a bug in applying Threshold (YOLOv8 models)
- Fixing a bug that crashes the tool when creating a polygon manually in image mode
- Fixing a bug that crashes the tool when deleting a polygon manually in image mode
- Flattening the Segmentation list in export COCO format in video mode
- `confidence` is changed to `score` in export COCO format to match the [standard Result format](https://cocodataset.org/#format-results)
- Fixing Image Resize before input to YOLOv8
- Fixing a bug that crashes the tool when opening merge models dialog with no image opened
- Removing YOLOv8 from the list of models in the merge models dialog (YOLOv8 is not supported in this feature)
- trying to save to Unspecified Export Path will show an Error Message now instead of successfully exported to Uknown path
- Removing Video Controls from image and dir mode (and vice versa)
- Fixing a bug that crashes the tool when opening a tool without any models, we added yolov8n
- Fixing a bug in export video shows all classes outside COCO as `toothpaste`
- Fixing a bug allows Export (Annotations and Video) when tracking is running
- Fixing a bug in play/pause button in video mode makes mouse clicks not working
- Fixing a bug in interpolation with keys (batch mode)
- Fixing a bug in interpolating first Frame
- Fixing a bug shows Downloaded for models that are not downloaded completely
- Navigating through tracked frames no longer creates a deepcopy of all shapes
- Downloading a model will show it immediately in merge menu (no need to Restart the tool)
- Non determined IDs have negative numbers assigned to them (starting from -1)
- Starting the tool with no models installed no longer crashes the tool
- fixing a bug in the model explorer hides some columns by default

New Features 🌟
- Adding Model Explorer to navigate, download, and select models
- Adding Merge models features
- Adding Runtime data (GPU Type, Memory)
- New README
- Updated requirements.txt
- fixing large git .pack files

Bug Fixes 🐞
- Fixing a bug when running YOLOv8models
- Fixing a bug causes ID Switch in Object tracking with YOLOv8