A Vision Transformer (ViT) is a specialized machine-learning model designed to process and classify image data. In mineral exploration, ViTs analyze geospatial data to help predict and assess the quality of mineral deposits. They work by embedding, encoding, and classifying input features (such as images or spatial data) to make predictions for each patch or grid of pixels.
Choosing the appropriate ViT model in the Embed Visual Features step is essential for optimizing predictions. The model chosen should be tailored to the specific mineral system you are targeting. For example, different models are available for porphyry copper or gold deposits, each trained with data related to those types of mineral systems.
Available Vision Transformer Models
Arch_IntrusionRelated_AuCu__c1p0.pt
Targeted for Archean intrusion-related gold and copper deposits.
Deposit Type: Archean intrusion-related deposits
Target Commodities: Au, Cu
VMS_Poly__c2p9.pt
Designed for identifying VMS deposits, targeting gold, copper, and zinc.
Deposit Type: Volcanic Massive Sulphide (VMS) deposits
Target Commodities: Au, Cu, Zn
Pegmatite_LCT_c0p2.pt
Tailored for locating lithium deposits in pegmatite formations.
Deposit Type: Pegmatite LCT-hosted lithium deposits
Target Commodities: Li
Epithermal_Au__c1p1.pt
Focuses on identifying epithermal-style gold deposits.
Deposit Type: Epithermal-hosted gold deposits
Target Commodities: Au
Epithermal_Poly__c1p0.pt
Deposit Type: Epithermal poly-metallic
Target Commodities: Au-Ag-Cu-Pb-Zn
Orogenic_Au__c4p0.pt
Built to target orogenic gold deposits using geological data.
Deposit Type: Orogenic gold deposits
Target Commodities: Au
Porphyry_CuAuMo_c2P1.pt
Specialized for porphyry systems targeting gold, copper, and molybdenum.
Deposit Type: Porphyry-related deposits
Target Commodities: Au, Cu, Mo
Carlin_Au_c1p0.pt
Focuses on Carlin-style gold deposits using predictive data.
Deposit Type: Carlin-style deposits
Target Commodities: Au
UtraMafic_CuNiPGE__c0p9.pt
Deposit Type: Ultramafic hosted Cu-Ni-PGE
Target Commodities: Cu-Ni-Co-Pt-Pd
Master_model.pt
Trained on a broad dataset for general exploration purposes. This model is designed to work across various systems and can deliver robust predictions even without specific training data.
Recommended Usage: Use when there is no commodity-specific model available for the targeted deposit type.
Deposit Type: General model (not specific to any deposit type)
Target Commodities: None (generalized)
None
Trains a model from scratch using only the data within the selected Area of Interest (AOI).
Recommended Usage: Use to train a model on only proprietary data, omitting VRIFY’s database.
Deposit Type: None (no predefined model)
Target Commodities: Based on the data provided in the AOI
Still have questions?
Reach out to your dedicated VRIFY AI Contact or email Support@VRIFY.com for more information.