A Graduate Student’s Summer 2026 Experience with a USDA-funded project: Collecting Field Images and Gaining Insights into Hardwood Bucking Practices at Virginia Sawmills

Project Overview

During Summer 2026, I had the opportunity to work on Project AgriProspect, an interdisciplinary project at Virginia Tech exploring how artificial intelligence (AI) could support hardwood log assessment and bucking decisions. The project is led by Dr. Pipiet Larasatie, with Dr. Chris Thomas leading the Computer Science component.

My summer assignment sounded simple at first: collect field images of Red Oak, White Oak, and Yellow Poplar logs.

The images would support the computer science side of the project, where AI models are being developed to recognize log characteristics from field images. I was not working alone. . Gede Abdi Dharma Pribadi, another Virginia Tech student, joined me for much of the fieldwork. On the computer science side, I worked with Md. Atabuzzaman, a Ph.D. student in Computer Science on Dr. Thomas's team. Dr. Thomas also provided technical guidance on what the computer vision team needed from the images we collected.

At first, I thought the main challenge would be taking enough photographs. I soon learned that field image collection involved much more than that. We needed access to operating sawmills, enough examples of our target species, consistent images, and enough data to account for the imperfect conditions that naturally come with fieldwork.

What began as an image-collection assignment eventually became a broader learning experience for me. Moving between Virginia Tech and sawmills across Virginia gave me an opportunity to see how research, Extension, industry relationships, and computer science can come together around a practical problem.

Building Industry Connections

One of the first things I learned was that fieldwork like this depends on relationships.

Dr. Brian Bond and Dr. Scott Barrett helped us make our first connections with hardwood companies. Tom Inman, president of Appalachian Hardwood Manufacturers, Inc. (AHMI), helped identify additional mills, while Ron Jenkins of the Virginia Loggers Association (VLA) reviewed our industry list and helped us consider other potential contacts.

Through these connections, I visited several hardwood operations across Virginia, including Turman Lumber in Christiansburg, Ontario Hardwood in Keysville, W.R. Deacon in Lexington, Campbell & Rodgers Lumber in North Garden, Blue Ridge Lumber in Covington, and Smith Mountain Land & Lumber in Huddlestone.

Our field locations developed through a practical combination of industry connections, access to operating mills, travel feasibility, and opportunities to photograph our three target species under actual production conditions. 

Rather than working in a controlled setting, we wanted to collect images of hardwood logs being handled and processed.

That meant spending quite a bit of time on the road. But the travel became part of the experience. Each mill had a different layout, equipment, and production environment. I went there primarily to photograph logs, but being inside these facilities allowed me to see much more of the hardwood industry than I could have seen from campus.

_{Video 1. Galit Prakosa and Gede Abdi Dharma Pribadi collecting hardwood log images at Blue Ridge Lumber during their Summer 2026 fieldwork for Project AgriProspect.(Permission was granted by Blue Ridge Lumber)}

Learning From the First Round of Fieldwork

One of my early image-collection visits was to Ontario Hardwood in Keysville.

I returned with photographs of all three target species and initially thought the trip had gone well. Then we looked at the images from the computer vision perspective.

Some of them were not quite what the project needed.

I had taken many photographs, but I began to understand that collecting images for computer vision required more than simply finding the right species and taking pictures. The computer science side of the team needed us to consider how the logs were positioned, which parts of the logs were visible, and whether we had enough images to work with, given that some could not be used.

After receiving a more specific image-collection protocol from Dr. Thomas's team, I returned to Ontario Hardwood and repeated the image collection using the new guidance.

That first experience changed how Gede and I approached the rest of our field visits. We became more deliberate about camera position and consistency, but we also tried to collect as many images as possible whenever we had access to suitable logs.

Field conditions are rarely perfect. A log may be partially covered, an angle may not capture what we expected, or some images may eventually be less useful for analysis. Having more images gave the computer science team more options and reduced the chance that missing or unusable images would create gaps in the dataset.

From that point forward, our goal was not simply to come back from a mill with photographs. We wanted to collect a large, useful set of images to better support the next stage of the project.

_{Figure 2. Field activities at Ontario Hardwood in Keysville, Virginia, provided opportunities to collect images of the target hardwood species and gain firsthand exposure to log handling and bucking operations (Permission was granted by Ontario Hardwood)}

Seeing More Than Logs

As I visited more mills, I started paying attention to things beyond the photographs I was there to collect. Each operation had its own equipment, production flow, workforce experience, and available resources.

At W.R. Deacon in Lexington, for example, I observed laser-assisted technology being used as part of the operation. 

It was different from the AI-assisted bucking concept we were exploring, but seeing technology already working in a mill made me think about our project differently.

On campus, it is natural to ask what a technology can do. In the field, I found myself asking another question:

Where would this technology actually fit?

Moving from one mill to another showed me that there may not be one answer. Each operation has its own characteristics, and its ability to explore or adopt new technology depends partly on the resources, infrastructure, and people already in place. What works well in one mill may need a different approach in another.

_{Video 2. Field activities at W.R. Deacon included students collecting images of hardwood logs and observing the company's processing operations. (Permission was granted by W.R. Deacon)}

Learning Extension from the Field

That realization also changed how I thought about Extension.

Before this summer, I tended to see Extension mainly as bringing university knowledge to industry and communities. Visiting the mills showed me that the connection works both ways. I went into the field with an AI project and an image-collection protocol, but I came back with a better understanding of the environments where that research might eventually be used.

The conversations during these visits were not formal research interviews, and I do not treat them as research findings. For me, they were part of learning how to listen and understand industry in its own setting.

As a graduate student, I began to see Extension less as simply bringing answers from the university and more as connecting university research with the experience, questions, and realities found in the field.

Connecting the Experience to My Research and Future Work

The sawmill visits also connected unexpectedly with my doctoral research on wood-residue utilization and valorization. While I was there to collect log images, I naturally noticed what happened to sawdust, chips, bark, lower-quality logs, and other materials beyond the mill's primary products. These observations were not dissertation data, but they gave me additional context for questions I am already asking in my research: Where do residues go? How are they used? And where might additional value be recovered?

There is also a longer-term reason this experience matters to me. I am a university faculty member in Indonesia, and I plan to return there after completing my Ph.D. I cannot simply take what I have seen in Virginia and apply it directly in Indonesia. The species, mills, resources, markets, and industry needs are different. But I can carry the approach with me.

I hope to create more opportunities for students to learn in manufacturing facilities, develop training programs that address real industry needs, and build stronger connections between universities and industry. The setting may be different, but the basic idea travels well: go to the industry, listen first, and build research or training around problems that matter there.

Learning Beyond the Sawmill

My summer experience also took me into an area that was new to me. Through Dr. Chris Thomas, I was introduced to Advanced Research Computing (ARC) at Virginia Tech and some of the computing resources available for research. Coming from a wood products background, I had little experience with this side of the process. It helped me see that collecting images in the field was only one part of a much larger interdisciplinary project.

More Than an Image-Collection Project

Looking back, this summer became much more than an image-collection assignment. Project AgriProspect took me from campus to sawmills, connected me with people outside my own discipline, and gave me a different way of thinking about how university research connects with industry.

More importantly, the experience changed how I understand Extension. I came to see it not simply as bringing knowledge from the university to industry, but as a two-way connection, one that requires us to spend time in the field, listen, and bring what we learn back into our research and education.

I started the summer with a camera and a list of three hardwood species.

I finished it with a broader understanding of what interdisciplinary research and Extension can look like at Virginia Tech.