Hello,
I was watching your webinar “Build Visual AI Agents with Generative AI and NVIDIA NIM”. My boss specifically asks whether the visual agent would be capable to detect a water leakage similar to the attached video. What prompt would it be necessary to put to the interface. Is general prompt like: “Is there a dangerous situation” enough or would I have to ask specifically: “Is there a water leakage” ?"Please let me know. And what would be the form of alert that we would get from the agent?
Hello! When testing VIA using a general prompt such as:
“Write a concise and clear dense caption for the provided warehouse video, focusing on irregular or hazardous events. Start and end each sentence with a time stamp.”
We receive the following response: 0.066666666:53.40: Water leak detected on warehouse floor and ceiling, accumulating near pallets and dock door area, and dripping from pipes.
While you may not need to give examples of what you want it to look out for, it is helpful to provide as few examples. Alerts from the agent will come in the form of text notification responses.
That sounds good! Is there a simple tutorial that will help me to begin with this kind of a NIM agent?
And also a tutorial for beginners with 3D assets generating NIM agent? We need to generate machines and equipment for our shop floor 3D digital twin. Then I would also appreciate a tool to generate furniture for offices and nature around our facility.
To get started creating agents using VIA, please sign up for the developer preview using a business email. Once you setup VIA, there are multiple examples that you can run in the documentation.
For your questions regarding 3D assets, I recommend you look into NVIDIA’s Omniverse.
I did what you proposed and applied for the VIA developer preview. Since I have some serious questions in regard of data security I tried to fill out the form to get in touch with Nvidia sales. Is this the right form? I still have no answer. I posted the form last Wednesday. I don’t know. Maybe it takes time to get the answer.
Did you ever get in touch with someone from sales? Would you mind posting your questions in the forum? If not, you can send me a direct message and I can get you the answers.
No, I have never been in touch with any sales employee from Nvidia. In general, I don’t mind sharing the issue. If I have to post some details, then I might be necessary to post them to you or sales directly. So, we need to know how Nvidia secure any video files we will upload for the detection tests or any other data we will provide. Since we’re certified acc. ISO 27001 we cannot afford to unveil the production space and any equipment, machines or products. The terms of use for VIA also allow Nvidia to use the data for Nvidia promotion. All data exposure to public is possible only over a written consent from our top manager. I’ll be probably forced to send you our non-disclosure agreement anyway.
When you launch VIA and its dependencies on your hardware, any videos uploaded for inference will remain locally and no data is shared with NVIDIA. If you use a 3rd-party endpoint for your VLM or LLM, then data is transmitted to that 3rd-party. For instance, if you use GPT4o for your VLM, then those video frames are shared with OpenAI. If you chose to use a locally deployed VLM such as VITA2 or VILA1.5, all data is kept locally and not shared through the VLM. If you have a sensitive question, feel free to message me directly.
which obviously succeeded because I got following message:
WARNING! Using --password via the CLI is insecure. Use --password-stdin.
WARNING! Your password will be stored unencrypted in /home/koule/.docker/config.json.
Configure a credential helper to remove this warning.
Login Succeeded
permission denied while trying to connect to the Docker daemon socket at unix:///var/run/docker.sock: Post “http://%2Fvar%2Frun%2Fdocker.sock/v1.47/images/create?fromImage=nvcr.io%2Fmetropolis%2Fvia-dp%2Fvia-engine&tag=2.0-dp”: dial unix /var/run/docker.sock: connect: permission denied.
Hi @jiri_mikulka , For the convenience of others to refer to, we try to discuss one related issue in a topic. Could you file a new topic about the new issue? Thanks
