TagScribeR v0.2.0-alpha

In this release basic Image Editing features are implemented and some bug fixes were done in Gallery.py, integration and use of OpenCV for image manipulation and display was done. Plans for a hotfix or update to migrate from QT/PIL for image display and adjustment in Gallery and Auto Captioning are in the works as OpenCV offers better performance and handles images much better.

TagScribeR v0.1.0-alpha

This is the v0.1.0-alpha release, currently the only full featured portion of the tool is the gallery widget.

Features include:

• Load Dataset directories for easy viewing and editing

• Tagging system with tag button creation and deletion and multi-select support for adding tags quickly to multiple images' captions

• Multi-select Images, Select All/Deselect All

• Dataset Collection creation/loading/deletion/management/copy selected files into dataset

• Search/Filter by tag

• Slider to change thumbnail sizes display

• Save Manual Edits button to save all manual edits

• Undo button - (a simple undo button that reverts any unsaved edits to any actively selected items. Note: This will only undo the last edits/tags done/added to the active selections, I could only implement a simple undo button due to my lack of extensive coding abilities)

• Button to open Collections folder in main TagScribeR root folder for convenience

• Persistent tag buttons and collections folders display between sessions

• Tag Sharing/Database - (In conjunction with making tags persistent all User created tag buttons are stored in a txt file within the TagScribeR root folder, making it easy for manual additions and sharing of User Created Tags database)

• Error logging and handling and exception handling

These are just a list of the current Gallery Widget's features, they can and likely will be expanded and improved upon in the future.

TagScribeR v0.01.0-alpha

This is the current build I have going for TagScribeR the UI does not match my mock up image at the moment and I'm slowly fleshing out all the features bit by bit. Currently this build is bare bones and does not have BLIP-2 Auto Captioning implemented, nor does it have the tags panel or collections/category panel. This release is to further my proof of concept and to show that I am currently developing things the best I can on my own. I'm determined to do this whether on my own or with help, but I would really love any help that any skilled/learned people may be willing to offer.

Current Build is meant for Windows OS and Python as that is what I am working with.

Current features:

• Load Directories and display/edit associated text files in Gallery tab/widget
• Load Directories with blank text fields meant for populating with Auto-Captions once BLIP-2 is integrated

To do list:

• Image selection (Multi-Select images functionality)

• Tag panel (collapsible tag panel with text field to add/create new tag buttons)

• Tag Buttons (functions to create tag buttons and functions to append current text/captions to selected images when clicking a tag button)

• Sort/Filter function (Sort/Filter by tags with displayed images & text being updated in the window to display only the images/texts associated with the tag being searched/filtered for)

• BLIP-2 integration (Integrate Blip-2 and Blip-2 settings to auto generate captions to text fields and save generated captions to new text files named after target images)

• Image Croppping/Resizing (Add image Cropping/Resizing widget and functions to crop images based on pixel dimensions and/or standard ratios, and resizing images based on pixel dimensions and/or ratios

• Tag Sharing (a tag database creation widget that allows you to save new tags to a json/database, load shared jsons into your already established database of tags and by extension load/hooking them to the tags panel in gallery so that new functional tag buttons are created from the shared tags you've loaded, collection/organization and overall management of tags databases/jsons) (Note: I'm unsure if json is the best approach, it's just the method I assume would be easiest to implement at first)

• Settings (A widget/tab for users to customize the look/style/feel of the UI, theme/color/etc, and hardware settings where the user can choose what backend settings the program will use for their Blip-2 functions CPU Only/Nvidia CUDA/AMD Directml, other settings as necessarry or as things evolve) (Note: I am on a Win 11 AMD 7900xt machine, so I might not have the hardware to develop the CUDA backend on my own)

Installation:

Git clone to your desired location

create new venv within TagScribeR folder

pip install -r requirements.txt

Running the Program:

.\venv\Scripts\activate

python main.py