Foaming and bubbles
Visualize candidate changes in foaming, air entrainment, and surface behavior.
Industrial Fluid AI
Turn existing cameras into fluid sensors.
ABYSS surfaces candidate changes in foaming, oil-film-like surface patterns, leaks, turbidity, liquid-level changes, and flow anomalies from ordinary camera footage. It supports 24/7 visual monitoring for fluid scenes that are hard to instrument with point sensors alone.
24/7
Monitoring
AI
Fluid Vision
PoC
Ready
Detection Scope
What ABYSS Detects
ABYSS focuses on visual fluid changes that are difficult to capture with point sensors alone.
Oil films and turbidity
Extract candidate changes in color, opacity, surface reflection, and oil-film-like patterns.
Leaks and overflow
Monitor visible wet areas, liquid spread, overflow events, and unexpected liquid movement.
Liquid-level changes
Track visible changes in tanks, channels, pits, and open water surfaces.
Flow anomalies
Visualize changes in speed, direction, mixing, agitation, and process stability.
Operator visual knowledge
Convert expert visual checks into consistent, auditable AI monitoring.
Capability Boundary
What ABYSS can do, and what needs validation.
ABYSS is not a universal image AI. It is designed for video of fluids such as water, oil, and chemical liquids, and it turns visible state changes into candidate signals operators can review.
check_circle
What ABYSS can support
- •Visualizing changes in flow direction, strength, and surface speed
- •Extracting candidate signals for foaming, oil-film-like changes, turbidity, dripping, and liquid-level changes
- •Recording abnormal scenes and supporting rewind review
- •Supporting patrol reduction, remote confirmation, and operator first checks
- •Checking fit against site standards through a PoC
- •Connecting alerts to existing systems, signal towers, email, Teams, Slack, or dashboards
rule
What needs validation first
- •Very small objects or regulated release-decision workflows
- •Calculating physical quantities such as flow rate, concentration, or material properties
- •Environments where the camera shakes heavily
- •Unstable lighting, reflections, viewing angle, or resolution
Depending on site conditions, additional cameras, lighting, external sensors, downstream models, or system integration may be required.
Technology
Fluid-centric architecture.
01
Physics-informed understanding
ABYSS is designed to recognize fluid motion itself, not only static visual patterns. It evaluates surface behavior, flow, reflection, and temporal change.
02
Real fluid video data
Models are tuned for the complexity of actual industrial scenes: lighting variation, camera noise, mixed liquids, foam, films, and moving surfaces.
03
Existing infrastructure
ABYSS can start from existing camera footage and connect alerts to dashboards, API outputs, and operational review workflows.
Use Cases
Industry Use Cases
Chemical & Materials
Chemical and materials plants
Monitor foaming in reaction tanks and visible leak candidates around pipes and process equipment.
#foaming
#leak-detection
Petroleum & Energy
Energy and refinery sites
Surface oil-film-like changes in wastewater pits, turbidity changes in cooling water, and visual signs of environmental risk.
#oil-film
#turbidity
Pharma & Bio
Pharmaceutical and bio processes
Support operator first checks and visual records for liquid level, bubbles, foaming, and filling-process changes.
#culture-monitoring
#liquid-level
View pharma filling use casearrow_forward
Water & Disaster
Water infrastructure and disaster monitoring
Monitor rapid water-level rise, debris accumulation, unusual ripples, and remote safety conditions.
#water-monitoring
#flood-risk
View river monitoring use casearrow_forward
Deployment
How ABYSS works
1
Connect existing cameras
Camera streams are connected to ABYSS in the cloud or through an edge deployment.
2
Analyze fluid state
The fluid AI model analyzes motion, shape, brightness, flow, and surface changes frame by frame.
3
Surface anomaly candidates
Visual deviations are converted into alerts while noise filters reduce unnecessary notifications.
4
Report and integrate
Detection data is visualized in dashboards and can be connected to existing alert or SCADA workflows.
4
Main Industries
6
Detection Types
24/7
Monitoring
PoC
Video First
Leadership
Founders
ABYSS is developed by a team combining industrial AI implementation, fluid AI model development, and new-business execution.
Kenyu Shimono
CEO
Completed graduate studies in Integrated Systems at Tokyo Institute of Technology.
Provided overseas expansion consulting for Japanese companies in Germany.
After returning to Japan, supported new-business consulting for major trading, electronics, and medical device companies.
Managed multi-site data infrastructure projects across departments in large enterprises.
Kengo Watanabe
CTO / AI Engineer
Completed graduate studies in Medical and Dental Engineering at Tokyo Medical and Dental University.
Worked on RPA and plant simulator development at Mitsubishi Gas Chemical.
Served as lead engineer at fluid AI startup AnyTech, developing proprietary fluid AI models.
Continued the fluid AI work through the subsequent M&A with JFE Engineering.
Evaluate ABYSS for your site.
Download the product deck or schedule a discussion about your current inspection challenge.