# US 12,008,699 B2 · Systems and methods for automating conversion of drawings to indoor maps and plans > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,008,699 B2 - Document ID: US12008699B2 - Status: Granted - Kind: B2 - Application number: 18/451,277 - Filed: 2023-08-17 - Published: 2023-12-07 - Granted: 2024-06-11 - Earliest claimed: 2022-03-10 - Assignees: Pointr Limited - Applicants: Pointr Limited - Inventors: Ege Çetintaş; Melih Peker; Umeyr Kiliç; Can Tunca - Web page: https://patentgrep.com/patent/US12008699B2 ## CPC - G06T 11/23: using straight lines or curves - G06F 30/12: characterised by design entry means specially adapted for CAD, e.g. graphical user interfaces [GUI] specially adapted for CAD - G06F 30/13: Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads - G06F 30/27: using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model - G06N 3/045: Combinations of networks - G06N 3/0475: Generative networks - G06N 3/094: Adversarial learning - G06N 20/00: Machine learning - G06T 11/20: Drawing from basic elements - G06T 11/65: on geographic maps - G06F 2111/12: Symbolic schematics - G06N 3/088: Non-supervised learning, e.g. competitive learning - G06N 5/02: Knowledge representation; Symbolic representation - G06N 5/04: Inference or reasoning models - G06N 20/20: Ensemble learning - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract Automating conversion of drawings to indoor maps and plans. One example is a computer-implemented method of creating an indoor map from a CAD drawing, the method comprising: preprocessing an original CAD drawing to create a modified CAD drawing, a text database containing text from the original CAD drawing, a CAD vector-image of the modified CAD drawing, and a CAD raster-image of the modified CAD drawing; creating a floor-level outline; sensing furniture depicted on the floor, the sensing creates set of furniture entities; identifying a room depicted in the CAD drawing; and creating the indoor map for a floor using the floor-level outline, the room-level outline and the room identity. ## Claims ### Claim 1 A computer-implemented method of creating an indoor map from a CAD drawing, the method comprising: preprocessing, by a device, an original CAD drawing to create a modified CAD drawing, a text database containing text from the original CAD drawing, and a CAD vector-image of the modified CAD drawing; identifying, by a device, a room depicted in the CAD drawing by: applying the CAD vector-image to a first machine-learning algorithm to produce a room-level bounding line and a first probability distribution regarding identity of the room; applying the room-level bounding line and the text database to a second machine-learning algorithm to produce a second probability distribution regarding identity of the room; applying the first and second probability distributions to a third machine-learning algorithm to generate a room identity; and selecting, based on the room-level bounding line, a room-level outline; and creating, by a device, the indoor map using the room-level outline and the room identity. ### Claim 2 (depends on 1) The computer-implemented method of claim 1 further comprising: sensing, by a device, furniture depicted within the CAD drawing, the sensing creates a set of furniture entities; and wherein applying to the third machine-learning algorithm further comprises applying the set of furniture entities to the third machine-learning algorithm to generate the room identity. ### Claim 3 (depends on 1) The computer-implemented method of claim 1 wherein identifying the room further comprises identifying each room of a floor depicted in the CAD drawing, each room having a respective room-level bounding line and a respective room identity. ### Claim 4 (depends on 1) The computer-implemented method of claim 1 wherein selecting the room-level outline further comprises: creating, by a device, an entity-based bounding line by applying the room-level bounding line to an entity-based post processing algorithm, the entity-based post processing algorithm creates the entity-based bounding line; creating, by a device, a graph-based bounding line by applying the room-level bounding line to a graph-based post processing algorithm, the graph-based post processing algorithm generates the graph-based bounding line; creating, by a device, a GAN-based bounding line by applying the room-level bounding line to a GAN-based post processing algorithm, the GAN-based post processing algorithm generates the GAN-based bounding line; and choosing, by a device, only one of the entity-based bounding line, the graph-based bounding line, or the GAN-based bounding line be the room-level outline. ### Claim 5 (depends on 4) The computer-implemented method of claim 4 wherein choosing the room-level outline further comprises: evaluating, by a device, the entity-based bounding line, the graph-based bounding line, and the GAN-based bounding line by a rule-based evaluator, the evaluating results in a rule-based selection; evaluating, by a device, the entity-based bounding line, the graph-based bounding line, and the GAN-based bounding line by a selection machine-learning algorithm, the evaluating by the selection machine-learning algorithm results in a ML-based selection; and choosing, by device, only one of the rule-based selection or the ML-based selection to be the room-level outline. ### Claim 6 (depends on 4) The computer-implemented method of claim 4 wherein creating the room-level bounding line further comprises: inflating the room-level bounding line to create an inflated bounding line; deflating the room-level bounding line to create a deflated bounding line; selecting a set of remaining entities based on drawing entities that reside between the inflated bounding line and the deflated bounding line; and creating the entity-based bounding line from the set of remaining entities. ### Claim 7 (depends on 6) The computer-implemented method of claim 6 wherein creating the entity-based bounding line further comprises: grouping drawing entities with respect to slope, the grouping results in a plurality of groups of drawing entities; and for each group of drawing entities, merging drawing entities that meet a predetermined distance criteria and a predetermined overlap criteria; the grouping and merging result in a set of remaining drawing entities. ### Claim 8 (depends on 4) The computer-implemented method of claim 4 wherein creating the graph-based bounding line further comprises: inflating the room-level bounding line to create an inflated bounding line; deflating the room-level bounding line to create a deflated bounding line; selecting a set of remaining entities being drawing entities that reside between the inflated bounding line and the deflated bounding line; and creating the graph-based bounding line from the set of remaining entities. ### Claim 9 (depends on 8) The computer-implemented method of claim 8 wherein creating the graph-based bounding line further comprises: grouping drawing entities of the set of remaining drawing entities with respect to slope, the grouping results in groups of drawing entities; for each group of drawing entities, selecting an outer drawing entity closest to the room-level bounding line, the selecting for each group of drawing entities results in a set of outer line segments; and extending at least one of line segment of the set of outer line segments; and wherein creating the graph-based bounding line further comprises creating the graph-based bounding line from the set of outer line segments. ### Claim 10 (depends on 4) The computer-implemented method of claim 4 wherein creating the GAN-based bounding line further comprises: creating, by a device, a first plurality of intermediate bounding lines by a first machine-learning algorithm; creating, by a device, a second plurality of intermediate bounding lines by a second machine-learning algorithm; creating the GAN-based bounding line by convolving the first plurality and second plurality of intermediate bounding lines. ### Claim 11 (depends on 10) The computer-implemented method of claim 10: wherein creating the first plurality of intermediate bounding lines further comprises applying a cropped raster image to the first machine-learning algorithm, the cropped raster image showing the room; and wherein creating the second plurality of intermediate bounding lines further comprises applying the room-level bounding line to the second machine-learning algorithm. ### Claim 12 A computer system comprising: at least one processor; a memory coupled to the at least one processor, the memory storing instructions that, when executed by the at least one processor, cause the at least one processor to: preprocess an original CAD drawing to create a modified CAD drawing, a text database containing text from the original CAD drawing, and a CAD vector-image of the modified CAD drawing; identify a room depicted in the original CAD drawing by causing the processor to: apply the CAD vector-image to a first machine-learning algorithm to produce a room-level bounding line and a first probability distribution regarding identity of the room; apply the room-level bounding line and the text database to a second machine-learning algorithm to produce a second probability distribution regarding identity of the room; apply the first and second probability distributions to a third machine-learning algorithm to generate a room identity; and select, based on the room-level bounding line, a room-level outline; and create an indoor map using the room-level outline and the room identity. ### Claim 13 (depends on 12) The computer system of claim 12: wherein the instructions further cause the at least one processor to sense furniture depicted in the original CAD drawing, the sensing creates a set of furniture entities; and wherein when the at least one processor applies the first and second probability distributions to the third machine-learning algorithm, the instructions further cause the at least one processor to apply the set of furniture entities to the third machine-learning algorithm to generate the room identity. ### Claim 14 (depends on 12) The computer system of claim 12 wherein when the at least one processor selects the room-level outline, the instructions further cause the at least one processor to: create an entity-based bounding line by applying the room-level bounding line to an entity-based post processing algorithm, the entity-based post processing algorithm creates the entity-based bounding line; create a graph-based bounding line by applying the room-level bounding line to a graph-based post processing algorithm, the graph-based post processing algorithm generates the graph-based bounding line; create a GAN-based bounding line by applying the room-level bounding line to a GAN-based post processing algorithm, the GAN-based post processing algorithm generates the GAN-based bounding line; and choose only one of the entity-based bounding line, the graph-based bounding line, or the GAN-based bounding line be the room-level outline. ### Claim 15 (depends on 14) The computer system of claim 14 wherein when the at least one processor chooses the room-level outline, the instructions further causes the at least one processor to: evaluate the entity-based bounding line, the graph-based bounding line, and the GAN-based bounding line by a rule-based evaluator, the evaluation results in a rule-based selection; evaluate the entity-based bounding line, the graph-based bounding line, and the GAN-based bounding line by a selection machine-learning algorithm, the evaluating by the selection machine-learning algorithm results in a ML-based selection; and choose only one of the rule-based selection or the ML-based selection to be the room-level outline. ### Claim 16 (depends on 14) The computer system of claim 14 wherein when the at least one processor creates the room-level bounding line, the instructions further cause the at least one processor to: inflate the room-level bounding line to create an inflated bounding line; deflate the room-level bounding line to create a deflated bounding line; select a set of remaining entities based on drawing entities that reside between the inflated bounding line and the deflated bounding line; and create the entity-based bounding line from the set of remaining entities. ### Claim 17 (depends on 16) The computer system of claim 16 wherein when the at least one processor creates the entity-based bounding line, the instructions further cause the at least one processor to: group drawing entities with respect to slope, the grouping results in a plurality of groups of drawing entities; and for each group of drawing entities, merge drawing entities that meet a predetermined distance criteria and a predetermined overlap criteria; the grouping and merging result in a set of remaining drawing entities. ### Claim 18 (depends on 14) The computer system of claim 14 wherein when the at least one processor creates the graph-based bounding line, the instructions further cause the at least one processor to: inflate the room-level bounding line to create an inflated bounding line; deflate the room-level bounding line to create a deflated bounding line; select a set of remaining entities being drawing entities that reside between the inflated bounding line and the deflated bounding line; and create the graph-based bounding line from the set of remaining entities. ### Claim 19 (depends on 18) The computer system of claim 18 wherein when the at least one processor creates the graph-based bounding line, the instructions further cause the at least one processor to: group drawing entities of the set of remaining entities with respect to slope, the grouping results in groups of drawing entities; for each group of drawing entities, select an outer drawing entity closest to the room-level bounding line, the selection for each group of drawing entities results in a set of outer line segments; and extend at least one of line segment of the set of outer line segments; and wherein when the processor creates the graph-based bounding line, the instructions further cause the processor to create the graph-based bounding line from the set of outer line segments. ### Claim 20 (depends on 14) The computer system of claim 14 wherein when the at least one processor creates the GAN-based bounding line, the instructions further cause the at least one processor to: create a first plurality of intermediate bounding lines by a first machine-learning algorithm; create a second plurality of intermediate bounding lines by a second machine-learning algorithm; and create the GAN-based bounding line by convolving the first plurality and second plurality of intermediate bounding lines.