Granted Patent B1

US 12,327,445 · App. 18/624,609 · Granted Jun 10, 2025

Inventors: Sven Eberhardt (Seattle, WA); Brian Westphal (Livermore, CA); John Bicket (San Francisco, CA)

Assignee: Samsara Inc.

G07C5/0825

Abstract

Claims (32)

1. A computerized method, performed by a user device having one or more hardware computer processors and one or more non-transitory computer readable storage device storing an inspection application executable by the user device to perform the computerized method comprising:

causing to display instructions on a display of the user device for the user to obtain inspection information of a vehicle, the inspection information including an inspection video of at least a portion of the vehicle or information related to the vehicle;

causing to determine a textual prompt including textual instructions to analyze the inspection video and to identify one or more particular vehicle components included in the inspection video;

causing to transmit a prompt to a large language model (“LLM”), wherein the LLM is a multimodal LLM capable of processing at least text and image data, wherein the prompt includes at least inspection information and the textual prompt;

causing to receive a first response from the LLM indicating that additional inspection information is needed to identify a particular vehicle component in the inspection video;

causing to display instructions on the display of the user device for the user to obtain the additional inspection information;

causing to transmit an updated prompt including at least a portion of the additional inspection information to the LLM;

causing to receive a second response from the LLM identifying a particular vehicle component of the vehicle;

causing to generate a component prompt including a plurality of possible inspection features that are specifically associated with the particular vehicle component;

causing to transmit the component prompt to the LLM; and

causing to receive a third response from the LLM indicating one or more inspection features of the plurality of possible inspection features that are identified by the LLM as associated with the particular vehicle component.

2. The computerized method of claim 1 , wherein said causing to transmit the prompt to the large language model is initiated in response to a user input indicating that inspection information has been obtained for analysis.

3. The computerized method of claim 2 , wherein the user input is provided via a hardware button of the user device or a software interface element of the inspection application.

4. The computerized method of claim 1 , wherein said causing to transmit the prompt to the large language model is initiated automatically in response to the inspection video being acquired by the user device.

5. The computerized method of claim 4 , wherein the inspection information includes a video stream that is periodically transmitted to the large language model.

6. The computerized method of claim 5 , further comprising:

analyzing the video stream with an image processing module to detect vehicle components in the video stream; and

extracting one or more still images from the video stream that include the detected vehicle component, wherein the one or more still images are included in the inspection information transmitted to the large language model.

7. The computerized method of claim 1 , wherein said causing to transmit the prompt to the large language model is initiated automatically in response to the inspection information being acquired by the user device.

8. The computerized method of claim 1 , further comprising:

updating an inspection report to include the one or more inspection features in associated with the particular vehicle component.

9. A non-transitory computer-readable medium storing a set of instructions that are executable by a user device having one or more processors, to cause the one or more electronic devices to perform a method, the method comprising:

displaying instructions on a display of the user device for the user to obtain inspection information of a vehicle, the inspection information including an inspection video of at least a portion of the vehicle or information related to the vehicle;

determining a textual prompt including textual instructions to analyze the inspection video and to identify one or more particular vehicle components included in the inspection video;

transmitting a prompt to a large language model (“LLM”), wherein the LLM is a multimodal LLM capable of processing at least text and image data, wherein the prompt includes at least inspection information and the textual prompt;

receiving a response from the LLM indicating that additional inspection information is needed to identify a particular vehicle component in the inspection video;

displaying instructions on the display of the user device for the user to obtain the additional inspection information;

transmitting an updated prompt including at least a portion of the additional inspection information to the LLM;

receiving a first response from the LLM identifying a particular vehicle component of the vehicle;

generating a component prompt including a plurality of possible inspection features that are specifically associated with the particular vehicle component;

transmitting the component prompt to the LLM; and

receiving a second response from the LLM indicating one or more inspection features of the plurality of possible inspection features that are identified by the LLM as associated with the particular vehicle component.

Assignments (1)

References Cited (400)

Cited By (16)

US 12,450,329 US 12,511,947 US 12,524,314 US 12,534,097 US 12,561,624 US 12,565,143 US 12,621,184 US 12,626,200 US 12,630,050 US 12,646,402 US 12,651,529 US 12,662,152 US 12,665,989 US 12,671,464 US 12,675,419 US 12,701,004