Unique Segmentation

Dependent Models

The outputs of the following models are used to determine any Unique Segmentation:

Model Name	Model Type
Image Preparation	Object Detection
Depth Registration	Semantic or Instance Segmentation
[Feature segmentation] e.g. Vein segmentation	Semantic or Instance Segmentation

Naming conventions for final Unique Segmentation products (the machine-learning model plus additional post-processing) will be as follows: Feature segmentation. For example, Quartz segmentation or Sulphide segmentation.

Data Processing

The outputs of the Datarock Unique Segmentation model when applied over depth-registered rows allow the area, percentage, count and polygon axis ratio of these segmented features to be calculated.

Below is an image illustrating the difference between a classification model and its potential outputs against an object detection, semantic segmentation (single-class) and instance segmentation (multi-class) model.

Detection of Unique Segmentation classes

Several segmentation classes can be detected by a Unique Segmentation model based on the training data collected during onboarding. An example of manually labelled examples of segmentation classes can be seen in the following image.

The following images show an example of segmented outputs for the vein classes as trained using the above manually labelled images. In the first row, the raw row image has been identified and cropped, the second row contains the predicted mask polygons for each of the 5 segmentation classes, and the third row includes these masks overlain across the raw cropped row with area statistics in cm².

Product Configuration Options

There are no configuration aspects to this product; however, the following post-processing logic can be applied depending on raw model results.

Smooth predicted masks
Fill gaps within predicted masks
Filter out smaller masks that don't form a vein
Merge predicted masks into another class
Create a new class based on predicted masked areas

Output Intervals

Default interval length: one row of core (~0.8m - 1m for standard core boxes)

Customisable interval available: Yes, this product can be run on any defined interval via uploading an assay or geology logging table to the Datarock Customer Success Team.

User Data

User data can not be uploaded to the Platform via CSV at the current time.

The following data is required for customisable intervals (assay or geology logs) to be sent to the Datarock Customer Success Team:

Hole_ID
Depth_From
Depth_To

Data Output

This class of models is processed by the Datarock team, and results can be downloaded from the Datarock Core platform.

Integration for running this product and viewing predictions in Datarock Core is ongoing.

The available CSV files include the following:

ProjectID_HoleID_segmentation_by_row.csv
ProjectID_HoleID_segmentation_by_log.csv
ProjectID_HoleID_segmentation_by_instance.csv

These CSV files contain the following headers:

File Header	Description	Row CSV	Log CSV	Instance CSV
hole_id	Customer’s Hole ID	Yes	Yes	Yes
depth_from_m	Start of interval	Yes	Yes	Yes
depth_to_m	End of interval	Yes	Yes	Yes
class_name	Predicted class name	No	No	Yes
area_cm2	Area of the individual predicted class	No	No	Yes
axis_ratio	Ratio of the height vs width for the individual predicted class	No	No	Yes
coherent_rock_sum_area_cm2	Total area of the row containing coherent rock pixels	No	Yes	No
incoherent_rockk_sum_area_cm2	Total area of the row containing incoherent rock pixels	No	Yes	No
total_rock_area_cm2	Total area of the row containing coherent and incoherent rock pixels	Yes	Yes	No
[unique segmentation class]_area_cm2	Total area of the row containing [unique segmentation class] pixels	Yes	Yes	No
[unique segmentation class]_area_%	Percentage of the row containing [unique segmentation class] pixels	Yes	Yes	No
[unique segmentation class]_count	Number count of [unique segmentation class] polygons identified within the interval	Yes	Yes	No
[unique segmentation class]_avg_axis_ratio	Average ratio of [unique segmentation class] polygon height to width across an interval	Yes	Yes	No
[uniqge segmentation class]_sum_thickness_cm	Total horizontal midpoint thickness of all predicted class polygons within an interval	No	Yes	No
[unique segmentation class]_avg_thickness_cm	Average horizontal midpoint thickness of predicted class polygons within an interval	No	Yes	No

Product Limitations

Limitations	Comments
Reliance on row detection and depth registration	The Unique Segmentation model is based on predicting and masking geological features within a row of drill core. The dependency on the depth registered rows being identified means that if a row is missed by the row model during Image Preparation, this row will not have the segmentation model applied.
Training is dependent on what can be seen within a row image	Datarock’s Unique Segmentation model relies on features being segmented using visually identifiable RGB features, some of which are too subtle or fine to predict from a photo, in particular if the resolution is poor. If segmentation classes are not identified, the resulting segmentation model will generally be lower than expectations based on any available site logging.
Training data must be representative of the whole area where segmentation is to be applied	If new imagery or a segmentation class is introduced to the model, the performance may decline as these examples were not trained during onboarding. An initial model evaluation will need to be undertaken to see the suitability of the model, in particular against any new imagery. Ideally, a new model version is trained to incorporate the new untrained drill core or segmentation class.

Document Version

Version	Date	Author	Rationale
1	29 June 2023	N Pittaway	Initial release
2	07 February 2025	N Pittaway	Added post-processing section
3	22 July 2025	N Pittaway	Naming convention plus minor edits
4	24 February 2026	N Pittaway	Update to Data Output table