Computer Vision for Research Computing

The Center for Research Computing (CRC) offers extensive expertise in applying advanced computer vision techniques to a wide array of research and development challenges. Computer vision enables computers to "see" and interpret images, videos, and other visual data. It involves developing algorithms and models that allow machines to identify, classify, and understand objects, scenes, and activities within visual information.  Numerous general purpose computer vision models already exist and domain or project specific models can be trained.

Our capabilities enable researchers to extract meaningful insights from visual data (or traditional numeric data transformed into image representations) This includes, but is not limited to, applications such as video analysis for membrane fabrication, disabled patient monitoring, rhizotron imagery to assess grasslands health, material defect detection, and automated quality control in manufacturing. We are happy to meet with students and faculty across disciplines to leverage cutting-edge computer vision methods for their R&D needs.  Simply email crcsupport@nd.edu to schedule a discussion. 

Bulk Image Captioning or Description Using Vision Models via Open WebUI API

If you have ever wondered about captioning a folder of images, or having a vision model describe all the images in a folder while still maintaining full control over the process, you can easily caption hundreds of images on the CRC-hosted Open WebUI platform with a single command—and let it run while you attend to other tasks. The code for this task is provided in both Shell and Python in the CRC User Documentation.

OpenAI Open-Weight Models

Earlier this month, on August 5, 2025, OpenAI officially released two open-weight models: gpt-oss-20B (lightweight) and gpt-oss-120B (requiring 80 GB of RAM). Both models are now available on the CRC-hosted Open WebUI platform

Welcome to Sign Up!

We invite you to sign up and explore the capabilities of our CRC-hosted Open WebUI GenAI platform at: https://openwebui.crc.nd.edu (accessible on campus or via VPN, Notre Dame Google login required). Currently, 128 users have registered on the platform.

Visual Studio Code (VSCode) is one of the most popular Integrated Development Environments (IDEs) and offers a convenient way to edit and manage code. However, when using VSCode on the CRC cluster, there are important performance and connection considerations to keep in mind.

This guide covers:

1. General Tips

VSCode can sometimes overload CRC’s front-end machines and file systems due to its background file monitoring. When you open a folder, VSCode periodically scans every file inside it using the rg (ripgrep) command. This is useful for version control and live updates, but:

To prevent overload:

Connection management:

2. Remote - Tunnels Tips

When you launch a tunnel, you’ll see a terminal link such as:

https://vscode.dev/tunnel/<tunnel_name>

This opens VSCode for the Web, which is suitable only for small projects due to feature and performance limitations.

Best practice for large projects: Use your locally installed VSCode and connect to the tunnel. This reduces server load. (Instructions are available in CRC’s documentation.)

3. Remote - SSH Tips

To further expand our reach, we’re launching a dedicated Google Group (rse_ai_asst_cop-list@nd.edu) to facilitate ongoing discussion, resource sharing, and knowledge exchange among participants. Whether you're a seasoned developer curious about AI assistance or a student just beginning to explore these tools, this community will provide a supportive environment to navigate the evolving intersection of artificial intelligence and research software engineering. Stay tuned for our first session announcement, and consider joining our mailing list to connect with fellow researchers who are shaping the future of computational research.

Name: Kristina Radivojevic

Department: Computer Science and Engineering

Advisor(s): Dr. Paul Brenner and Dr. Tim Weninger

Research Summary: The emergence of internet-facing Multimodal Foundation Models (MFMs) poses a potential direct threat to public discourse, as humans often fail to identify AI-generated impostors. While this technology could facilitate the manipulation of online users, my thesis, "Cyber Technology for Public Discourse," argues that its generative capabilities can also be leveraged for constructive ends. My research combines new cyberinfrastructure with the responsible use of MFMs to both assess the current state of public discourse and enhance it by mitigating toxic speech while protecting freedom of expression. To support this work, we developed the Public Discourse Sandbox (PDS), a secure platform for conducting human-AI and AI-AI discourse experiments that are not feasible on public social media. Finally, the goal of this thesis is to build a core conversational chatbot architecture that can be strategically tailored to meet the unique needs of different industries. 

Favorite Campus Spot: My favorite campus spot to study and research has always been the CRC office, quiet and full of supportive people, always happy to help. I also enjoy having long walks around the lake.

Every Wednesday and Thursday
2:00 – 3:00 p.m.
812 Flanner Hall (map)

The CRC offers multiple training opportunities for both new and existing users. We periodically provide short courses and other learning opportunities, which are advertised on our website and through email lists. In-person office hours are held every Wednesday and Thursday from 2:00-3:30 p.m. in Flanner Hall, room 812, on a first-come, first-served basis. You can also arrange a Zoom meeting at your convenience by emailing CRCsupport@nd.edu with your availability. We recommend bringing a laptop to in-person sessions.

For more information, please contact CRCsupport@nd.edu.