Illustrative future concept: a solar monitoring buoy floats above a fish moving through open water

AI VISION / THE FUTURE OF PHANTOMTASK

Let nature
keep moving.

Our vision: intelligent field systems that help us understand the living world with less interference.

A future direction for PhantomTask

Concept imagery. Buoy design and visual identification remain under development.

FROM FIELD TO CLARITY. AND BEYOND.

A natural
next chapter.

Today, we connect remote equipment to useful information. Tomorrow, we want those systems to understand more of what is happening around them.

Computer vision could extend the same PhantomTask approach: collect in the field, automate the work, and turn observations into insight.

Wildlife monitoring is one place we want to begin. It is part of a broader ambition to make difficult-to-monitor environments easier to understand.

PHANTOMTASK / CONCEPT 0110 CONNECTED COMPONENTS

Day preview · beacon offLocal time · night 7 PM–7 AM

Preparing the 3D buoy…

Explore the ten components and their roles below.

Drag to rotate · hover or select a part

Explore how the system fits together.

Ten connected components, from solar energy above the surface to imaging below. Hover to explore a part; select it to keep it in focus. The other components separate into clear, spaced positions.

Concept assembly · not a final engineering specification

FROM FOUNDATION TO FUTURE

Practical engineering.
A bigger horizon.

Remote collection, Python automation, equipment integration and dashboards are the foundation we’re building on.

This buoy concept brings that approach onto the water: power and communications above the surface, with a proprietary imaging system below.

Explore our existing approach

Bring the monitoring system
to the fish.

A buoy could become an observation point within a waterway, bringing cameras, computing and environmental context together while fish continue past.

Less handling.
More natural movement.
A better view.

We want to investigate where visual observation can reduce the need for capture, physical identifiers and repeated intervention.

PowerConnectivity

Computing at the observation pointProcess · store · transfer

WATER SURFACE

Underwater imagingEnvironmental context

Concept system, not a final hardware specification.

How this relates to established monitoring methods

Capture, handling, implanted or attached identifiers, specialized detection equipment and passage structures remain important scientific tools. We want to explore another option where images can provide useful evidence.

Camera placement, power, communications, GPS, above-water imaging and environmental sensors would depend on the location and research question. Visual methods would complement established techniques where they can be validated.

What if the fish itself
became the identifier?

We want to investigate whether natural characteristics could connect one encounter with another. A fish moving freely. An image worth keeping. A familiar pattern seen again.

VISUAL FEATURE RESEARCH

A marking. A pattern. A fish seen again.

Illustrative concept — sample data

Illustrative fish image used to explain research into natural visual features; not identification evidence

Concept imagery · no model is running

Markings & pigmentation

Could the arrangement of spots or coloration remain recognizable across encounters? Lighting and biological changes must be accounted for.

That possibility needs careful validation. Lighting, water clarity, viewing angle and changes in appearance can all affect a comparison.

One encounter.
A longer story.

From a moment in the water to a record that helps people ask better questions.

Observe freely.

Bring imaging into the habitat. Let fish pass naturally through a useful field of view.

A passing fish
A useful image
Environmental context

Recognize naturally.

Explore how species, markings and other visible characteristics could help us understand what the camera sees.

Detection
Species
Natural features

Connect encounters.

Investigate whether a new observation can be associated with a fish seen before, keeping uncertainty with the record.

An image
A time and place
A possible association

Understand more.

Bring observations into PhantomTask’s data workflows to explore activity, movement and habitat over time.

Searchable histories
Patterns over time
Useful dashboards

Repeated observations could help reveal species activity, migration timing and habitat use. Sightings would not imply continuous tracking between encounters.

Explore the proposed nine-stage workflow

CONCEPTUAL WORKFLOW

Observe

A buoy-mounted camera would continuously or periodically observe an appropriate section of water, allowing fish to pass naturally.

Potential output: An image or video sequence
Research question: Can we collect useful views without redirecting fish?

Illustrative concept — sample data. Real-time detection is an ambition; transfer and dashboard update frequency would depend on deployment conditions.

Observe. A buoy-mounted camera would continuously or periodically observe an appropriate section of water, allowing fish to pass naturally.

Detect. Computer vision would look for a fish entering the camera’s field of view and separate possible sightings from other objects.

Identify species. A model would estimate the likely species from visible characteristics, retaining uncertainty and allowing an unknown result.

Capture. The system would select useful frames with sufficient detail, rejecting views that are blurred, obscured or poorly lit.

Analyze features. Research would examine scale patterns, markings, pigmentation, body features, fins and other potentially repeatable natural characteristics.

Compare. Useful observations would be compared with historical records. Similarity could suggest candidates for review; it would not by itself establish identity.

Candidate record. Sufficient evidence could support an association with an existing candidate. Otherwise, retain an unassigned observation or a provisional record without assuming it is a new fish.

Record. Potential records include the image, species estimate, candidate ID, confidence values, date, time, monitoring location, direction of travel, observation history and environmental sensor readings.

Visualize. Appropriate information would transfer to PhantomTask dashboards: sightings, species activity, observed movement, migration timing, image and location history, model confidence, environmental conditions and equipment health.

Illustrative forest river flowing freely between natural rocks and trees

THE REASON TO BUILD IT

Better data.
Less disturbance.

Observe the ecosystem without unnecessarily changing the ecosystem in order to observe it.

Through our work with WSB, we see an opportunity to bring environmental expertise, field engineering and AI together.

Our ambition is to help people understand wildlife movement and behavior with less unnecessary interaction with animals and their habitat.

Illustrative underwater habitat with trout passing above a sunlit riverbed — A living environment. A lighter footprint.

For wildlife.

Explore observation with less capture, handling and physical identification.

For waterways.

Study movement and habitat while seeking ways to reduce disruption.

For people.

Make useful field information easier to revisit, compare and understand.

Explore the potential benefits and applications

Reduced animal handling

Explore where repeated capture and physical handling could be reduced.

Fewer invasive procedures

Investigate where visual methods could complement or reduce the need for implanted identifiers.

Less waterway disruption

Seek configurations that do not require restrictive passage structures.

Longer observation windows

Explore remote observation over longer periods, subject to field conditions and system availability.

Natural behavior

Study activity with less direct interaction between animals and monitoring personnel.

Image-backed records

Retain visual context alongside each usable observation.

Movement and migration

Explore patterns across repeated sightings without assuming continuous tracking between them.

Habitat understanding

Relate observations to location, environmental readings and habitat context.

Safer field operations

Potentially reduce repeat visits to difficult, isolated or hazardous monitoring locations.

Validated observations could inform habitat restoration, migration corridors, fish passage, conservation planning and infrastructure decisions. These are intended applications, not reported outcomes.

AI-generated habitat imagery illustrating the environments we hope to understand.

A vision takes people.
And work.

Good imagery. Careful research. Practical field engineering. We’re bringing these disciplines together to explore what comes next.

Training the next layer of understanding.

We’re using NVIDIA GPUs to train and develop computer-vision research, including detection, classification and visual recognition.

Research computing is separate from the eventual onboard hardware. No NVIDIA affiliation or endorsement is claimed.

See the proposed research and development path

A proposed sequence, with no completion dates or completed stages claimed.

Build the evidence

Dataset development
Collect, classify, clean and organize suitable fish imagery.
Fish detection
Train and evaluate models that detect fish in images and video.

Understand what we see

Species classification
Train and validate models for relevant species, retaining unknown results.
Visual feature research
Investigate which natural characteristics are repeatable enough for individual recognition.

Test identity in context

Individual re-identification
Research comparisons between new observations and historical records.
Environmental testing
Evaluate clarity, lighting, depth, motion, viewing angle, occlusion, seasons and biological appearance changes.

Move carefully into the field

Buoy prototype
Integrate imaging, computing, communications, sensors and power into a field-ready concept.
Field validation
Compare observations with scientifically established reference methods before relying on identification claims.

Make observations useful

Remote integration
Connect suitable records to PhantomTask’s remote data infrastructure.
Automated wildlife intelligence
Develop organized, searchable observation histories and long-term monitoring dashboards.

BUILD THE FUTURE WITH US

Help shape
what comes next.

We’re looking for people who understand the environments, the animals and the technology. Bring your expertise to PhantomTask’s next chapter.

Wildlife & aquatic research

Fisheries researchers, wildlife and marine biologists, aquatic ecologists, universities and conservation organizations.

AI & imagery

Computer-vision researchers, machine-learning engineers, dataset contributors and camera or imaging specialists.

Field systems & support

Edge-computing specialists, hardware developers, field-testing and pilot partners, research sponsors and environmental engineering organizations.

Help us build it Back to PhantomTask

Opens an email draft to Remote@biorngroup.com with the AI Wildlife Identification Project in the subject. Review and send it in your email app.

Let naturekeep moving.

A natural next chapter.

Practical engineering.A bigger horizon.

Bring the monitoring systemto the fish.

What if the fish itselfbecame the identifier?