Back to results
US 12,597,202 B2Granted
Geologically meaningful subsurface model generation based on a text description
AssigneeSchlumberger Technology Corporation
InventorAnatoly Aseev
Life of this patent
Estimating the term…
Abstract
A method, system, and computer program product are provided for generating a subsurface model by an image generating machine learning model (MLM). A text prompt is received describing a subsurface geological feature. A text encoder encodes the text prompt into a text embedding. The text embedding is processed by the image generating MLM, specifically trained on pairs of text descriptions and subsurface models, to generate subsurface models from text. The image generating MLM outputs a subsurface model comprising the subsurface geological feature that corresponds to the text prompt.
Claims23
A method for generating a subsurface model by an image generating machine learning model (MLM), comprising:
- receiving a text prompt describing a subsurface geological feature;
- receiving a latent seed comprising a noise vector;
- processing the noise vector to create a grid of latents, the grid of latents comprising a low-resolution image representation of the subsurface model;
- encoding, by a text encoder having frozen weights, the text prompt into a text embedding;
- processing the text embedding and the grid of latents by the image generating MLM specifically trained on pairs of text descriptions and existing subsurface models to generate subsurface models from text, wherein the image generating MLM modifies the grid of latents on the text prompt to generate conditioned latents that infuse a semantic meaning of the text prompt into the low-resolution image representation;
- performing, via the image generating MLM, a diffusion process using a scheduler algorithm configured to iteratively refine the conditioned latents, the performing the diffusion process including a reconstruction comprising a series of denoising steps that gradually reduce noise in the low-resolution image representation to shift the low-resolution image representation into a coherent image;
- and
- outputting, by the image generating MLM, the subsurface model comprising a model of the subsurface geological feature that corresponds to the text prompt, wherein the outputting the subsurface model comprises displaying the coherent image including variations in seismic properties of layers in the subsurface geological feature, the variations comprising the noise and a frequency.
5 more elements
A system comprising:
- a computer processor;
- and
- a non-transitory computer readable storage medium storing program code, which when executed by the computer processor, performs a plurality of operations comprising:
- receiving a text prompt describing a subsurface geological feature;
- receiving a latent seed comprising a noise vector;
- processing the noise vector to create a grid of latents, the grid of latents comprising a low-resolution image representation of a subsurface model;
- encoding, by a text encoder having frozen weights, the text prompt into a text embedding;
- processing the text embedding and the grid of latents by an image generating machine learning model (MLM) specifically trained on pairs of text descriptions and existing subsurface models to generate subsurface models from text, wherein the image generating MLM modifies the grid of latents on the text prompt to generate conditioned latents that infuse a semantic meaning of the text prompt into the low-resolution image representation;
- performing, via the image generating MLM, a diffusion process using a scheduler algorithm configured to iteratively refine the conditioned latents, the performing the diffusion process including a reconstruction comprising a series of denoising steps that gradually reduce noise in the low-resolution image representation to shift the low-resolution image representation into a coherent image;
- and
- outputting, by the image generating MLM, the subsurface model of the subsurface models generated from the text, the subsurface model comprising a model of the subsurface geological feature that corresponds to the text prompt, wherein the outputting the subsurface model comprises displaying the coherent image including variations in seismic properties of layers in the subsurface geological feature, the variations comprising the noise and a frequency.
8 more elements
A computer program product comprising a non-transitory computer readable storage medium storing program code, which when executed by a computer processor, performs operations comprising:
- receiving a text prompt describing a subsurface geological feature;
- receiving a latent seed comprising a noise vector;
- processing the noise vector to create a grid of latents, the grid of latents comprising a low-resolution image representation of a subsurface model;
- encoding, by a text encoder having frozen weights, the text prompt into a text embedding;
- processing the text embedding and the grid of latents by an image generating machine learning model (MLM) specifically trained on pairs of text descriptions and existing subsurface models to generate subsurface models from text, wherein the image generating MLM modifies the grid of latents on the text prompt to generate conditioned latents that infuse a semantic meaning of the text prompt into the low-resolution image representation;
- performing, via the image generating MLM, a diffusion process using a scheduler algorithm configured to iteratively refine the conditioned latents, the performing the diffusion process including a reconstruction comprising a series of denoising steps that gradually reduce noise in the low-resolution image representation to shift the low-resolution image representation into a coherent image;
- and
- outputting, by the image generating MLM, a subsurface model of the subsurface models generated from text, the subsurface model comprising a model of the subsurface geological feature that corresponds to the text prompt, wherein the outputting the subsurface model comprises displaying the coherent image including variations in seismic properties of layers in the subsurface geological feature, the variations comprising the noise and a frequency.
5 more elements
Dependencies are inferred from the claim text.