segmenteverygrain is a Python package that aims to detect sedimentary grains (or grain-like objects) in images, created by Dr. Zoltan Sylvester at UT Austin. Please see the original repo for details and guidance. This project is a fork of the original, and is in no way approved or supported by the creator(s), maintainer(s), or sponsors of segmenteverygrain.
This fork was independently created to support upcoming work by Dr. Emma Lodes at Arizona State University. It is not approved or supported by Arizona State University, any employees or representatives thereof, or anyone other than the creator of this fork.
This project is in active development. Utility, accuracy, and safety are not guaranteed. Use this code at your own risk.
This unofficial fork attempts to build on the core features of segmenteverygrain by improving user interactivity and adding auxiliary features.
The original automated grain detection and segmentation process as laid out in Segment_every_grain.ipynb has not been significantly modified except for how results are saved. The notebook itself has not been modified at all.
The original notebook (above) can be executed directly, but results will not be wholly compatible with the features below. Compatible results can be generated by running the automatic detection and segmentation process through the GUI, or by editing the run_auto.py script for appropriate input/output locations and running it directly. The script only implements the normal image processing method from the notebook, while the GUI includes both the normal and large-image methods.
The original manual editing interface has been modified to facilitate:
- Increased FPS, especially with high grain counts
- Using the same interface to create, delete, merge, and measure grains
- More control over grain detection, including a selection box and arbitrary numbers of foreground/background prompts
- A toggleable info box to display information about the last-selected grain
- Undoing the most recently-created grain
- Minor bugfixes like not resetting the zoom level when creating a grain
Run the new interactive interface through the GUI or directly using run_edit.py, editing it for appropriate input and output locations. Navigation within the interface is described in the matplotlib documentation. Additional controls are:
- Left click: Place foreground prompt or select existing grain
- Shift + left click: Create or adjust box selection tool
- Right click: Place background prompt
Escape: Remove all prompts and unselect all grainsc: Use selection box and/or foreground/background prompts to detect a graind: Delete selected (highlighted) grainsi: Toggle info box, showing details of the most recently-selected grainm: Merge selected grains (must be touching)z: Delete the most recently-created grain
The results of segmenteverygrain can be used to simulate the results of a manual grid count. This provides less information than segmentation of a complete image, but is more comparable to traditional methods. Currently there is no fine-grain cutoff threshold and points not directly landing on a grain are simply excluded.
After using segmenteverygrain to generate a grain info csv (ending in _grains.csv), generate a grid count through the GUI or by directly editing and running run_point_count.py.
The core functions of segmenteverygrain have been wrapped into an optional, experimental GUI in order to provide a familiar user experience. Features include:
- Automated grain detection and segmentation, mostly following the process laid out in the original Segment_every_grain.ipynb Jupyter Notebook
- Large-image grain detection and segmentation, again per the original notebook
- Manually-guided detection/editing of grains using the improved interactive interface
- Grid count simulation
- Unit conversion
The GUI was created using Kivy, which adds additional dependencies. Kivy and matplotlib integration (required) can be installed using pip install kivy kivy_garden.matplotlib --extra-index-url https://kivy-garden.github.io/simple/. Refer to the official Kivy site for more details.
Run the GUI from segment.py, use the buttons at top to load an image and/or a csv with grain data that was generated by another method, and access the desired functions from the toolbar on the bottom. If you want to modify the GUI itself, note that its layout is defined in segment.kv
This fork of segmenteverygrain is tentatively feature-complete for its intended usage, though questions and suggestions are always welcome via the issue queue.
Current work is focused on bugfixing and documentation, as well as upstreaming any features that the creator(s) of the original project might find useful.