BenchBot

High-Throughput Imaging, Custom-Built to Feed Agriculture’s AI Revolution

Agriculture is facing a data crisis. American farmers are grappling with pest resistance, water scarcity, degraded soils, and a shrinking workforce. Affordable site-specific precision technologies that boost productivity, cut input costs, and sustain rural economies aren’t a luxury; they’re a necessity.  Artificial intelligence can deliver those technologies, but AI needs data, lots of it. The training-data bottleneck is one of the biggest barriers to a digital agriculture revolution – and BenchBot is built to break it.

What Is BenchBot? 

BenchBot is an open-source imaging platform designed to generate the large, high-quality datasets needed to power agricultural artificial intelligence. Built on a modular framework, BenchBot semi-autonomously captures high-resolution plant imagery at scale, creating pre-labeled, AI-ready datasets that feed directly into computer vision workflows and machine learning model development. By automating daily image collection across crops, weeds, and specialty plants, 

BenchBot breaks one of the biggest barriers in digital agriculture: access to reliable training data. The result is a cost-effective, flexible phenotyping system that helps researchers, developers, and producers build smarter tools for real-world agricultural decision-making. Together with DASH’s broader digital infrastructure, BenchBot is helping move agricultural AI from limited proprietary systems into a shared public ecosystem that researchers and producers can actually use.

BenchBot & the Agricultural Image Repository

BenchBot is the primary engine powering the National Plant Agricultural Image Repository (AgIR) – a growing cloud-based collection of high-resolution agricultural imagery and metadata available to researchers, technology developers, and farmers alike.

Here’s how the pipeline works:  

Capture → Transfer → Process → Deploy

  1. Capture: BenchBot photographs plants at high resolution, collecting large volumes of consistent, structured imagery across growth stages and environmental conditions.
  2. Transfer: Large image files are securely delivered via Globus — a trusted, high-speed research data transfer platform — to USDA’s SCINet high-performance computing (HPC) cluster, with a mirrored cloud environment coming soon for non-SCINet users.
  3. Process: Imagery enters a computer vision pipeline where it is preprocessed, color-calibrated, and passed through semi-automatic annotation workflows. The result: clean, labeled, AI-ready datasets.
  4. Deploy: These datasets are available to train customizable machine learning models like ModCam for breeding research, production decision tools, and on-farm applications — all freely accessible to the public.

Together, BenchBot and the Agricultural Image Repository form an expanding public ecosystem of cutting-edge AI solutions, democratizing agricultural technology and powering affordable tools to help farmers make real-time, data-driven management decision

Become a BenchBot Partner

BenchBot is open for new collaborators. By hosting a BenchBot, you get more than a machine, you gain access to a customizable research platform supported by DASH technical expertise. 

Our team provides site planning, installation, training, routine service, and troubleshooting to keep your system running. In return, host sites contribute to the growing image repository and gain access to the full data pipeline and HPC processing environment.

Site Requirements

BenchBot’s gantry dimensions are customizable to your site. Universal requirements include:

    • A rectangular, open area to fit pots of target plant species

    • A perimeter barrier for plant and equipment security

    • Level, flat ground for BenchBot rails (≤3° slope in any direction; most surfaces are acceptable if anchors and rebar can be installed)

    • Irrigation system with appropriate drainage

    • Extra utility space at each end of the track

Sizing Needs

    • Gantry Width: 12 ft – 32 ft

    • Track Length: 60 ft – 200 ft

Resource
Requirement
Power
240V, 30A single-phase (or single leg of 208V 3-phase)
Internet
Sustained upload speed of ≥100 Mbps; 1 Gbps preferred
Water
Sufficient source for crop irrigation needs
Potting
Black plastic pots with peat/bark/perlite mix; slow-release fertilizer recommended

    • On-site staff available for day-to-day operations

    • SCINet accounts for users who need data access (required for SCINet-based data delivery; cloud mirroring coming soon)/

Whether you are a research institution, breeding program, public agency, or agricultural innovation partner, BenchBot is designed to fit your site and amplify your research. Reach out to learn how to bring BenchBot to your location.