Artificial Intelligence for the FISH FARM inspections

Customized image processing for ROV/camera
Objective and fast measurements
Shorter inspection time on each cage
Machine learning training / environmental data

*meet GREENER standards with AI solutions!

Supported by:

Used by:


Solution for fouling:

We are developing custom automatic AI analysis of underwater ROV/camera video materials to inspect the fish tanks fouling over-growth.

  • secure net-cleaning only on a need-to basis (instead of at fixed intervals as is the case today)
  • to provide objective measures
  • reduces costs (done only when needed),
  • improve animal welfare; secures sufficient oxygen levels for the fish and do as little disturbance for fish as possible
  • protects the environment (reduce the spill of copper based anti-fouling to the seabed)
    The primary effect: reduced use of anti-fouling, specifically copper-based resin, needed to clean the nets.

Demo solutions:


Measurement of daily fish mortality (bottom of the cage).

Dead fish is lying at the bottom of the cage. We detect each fish at each frame of the video. We time-trace each fish for the position in the cage and the counting. Colorful tracing lines (left) show relative position while tracing each dead fish for counting. The color of the tracing lines corresponds with the color of the marking on the fish (right)

Presence/detection of details on Salmon video

By tracing such a small and specific structure/feature on the foreground fishes on the video – like the fisheye, we demonstrated the ability to detect any other structures like a presence/detection of Salmon Lice’s (Caligus) or other illnesses or deformations.

Counting and tracing each fish

We detect and subsequently trace the new fish. We also give an estimated number of fish in optically dense fish clusters!

This example demonstrates the power of detection and tracing in time moving fish in the cage! Our AI-based solution brings an efficient method to count fish in the fish tank daily!

Salmon spot Uniq patterns detection

Analysing details on the selected fish; even in the deep and dark waters, KARMENstudio can distinguish the unique pattern of the spots on the salmon back, scale many details and estimate the biomass.


  • delivering fully automatic inspection image/video analysis and reporting.
  • All solutions and deliverables are customized,
  • observations and data analysis are presented in various formats – oriented on need-to-basis.
  • In some use cases, KARMENstudio provides automated reporting. We generate the inspection report alone and only alert the human when there is an anomaly detected.

KARMENstudio is built for scientific & industrial DATA

The range and volume of data that can be imported and exported are significant.

We fully support the following file types:

  • Full-range images, full-range video formats;
  • 360° images from various camera types 
  • Document file formats including .doc, .pdf, .xls, .csv, etc.
  • 3D models (like .OBJ or .DAE, etc)
  • Hyperspectral layers/markers of the same fields (up to 100 layers)

We combine AI with other techniques to solve a problem

  • AQUACULTURE assets are complex, and the real world is noisy!
  • We spend time understanding the roots of the problem!

We combine multiple approaches and techniques to solve complex problems.

We create:

  • Physical models
  • Sets of image processing approaches
  • Multiple approaches and techniques such as:
    • Computer vision
    • Geo-Positioning
    • Our own Artificial Intelligence solutions

Private Label

  • You’ll never be boxed in.
  • We are all about customization!
  • Dedicated AWS server for each customer


Fish farm inspections (ROV) can meet GREENER standards with AI solutions!

Fouling is the multitude of organisms growing on the fish farm nets...

read more

Machine learning training for Fish farm inspections (ROV) integrates extensive ROV operator’s experience into the AI solution!

Each problem has specific field knowledge - where years of experience and competencies are...

read more

Visualization of measured values for Fish farm inspections (ROV)

Fish farm assets are complex, and the real world is noisy...

read more

What KARMENstudio can help you to achieve

More objective and faster digital capture of your assets

Generate revenue and improve forecasting

Reduce the mean time and cost of repair

About us

"We are here to provide you AI ( machine learning ) tools for fast and objective image/video analysis!"

KARMENstudio team - committed to bringing solutions to help

We have a core team of 7 dedicated scientists, entrepreneurs, and developers. Led by:

Ana Bedalov

CEO Physicist

Saša Kijuk

Head Software Developer

  • 20 years of experience in image analysis
  • very skilled machine/deep learning developer

Tonći Kozina

Signal analysis engineer

  • 22 years experience in signal analysis, analytical operations, and development for military purposes.
  • rich experience in developing and maintaining software applications

Ana Bacelj


  • MSc astrophysics
  • very precize, selfdriven and creative image analysis specialist

Denis Zorić


  • 15 years of experience in logistic protocols and statistics
  • IP and transfer of technology in medical bionics
  • testing and verification of algorithms

Jelena Radić

Automatization engineer

  • with a fresh master in automatization engineering, she inherited all of our "problems" concerning full automatization of the processes and optimization of the usage of the server and human workers.  
  • Her pragmatic mind is conquering step by step our protocols and giving all of us a space to breathe!

KARMENstudio team -bringing custom smart structure recognition solutions!