ITP, InTech Partner

Can a camera continuously estimate the share of fine particles in a falling ore stream? 📹
That was the question behind this proof of concept.

The approach:
Detect the ore stream → separate particles at a 5 mm threshold → estimate the fine-particle share → compare it with the reference measurement.

We tested 3 segmentation approaches — binary, Otsu, and adaptive — across 3 camera types: Axis, Canon, and infrared.

The result: ~10.8% small fraction vs. 12% reference.

Have a measurable visual quality parameter and a reference method? We can structure a PoC around your camera samples.

🌐 ai.itp.biz

We’re heading to Porto for SAP Inside Track Porto 2026! 🇵🇹

On September 26, the ITP team will be at Forte de Gaia, exploring the latest developments across SAP Business AI, AI agents, SAP BTP, Business Data Cloud, and S/4HANA.

We’re looking forward to exchanging ideas, sharing our experience, and learning how these technologies are being applied in real-world projects.

Will you be there? Let’s meet in Porto!

📍 Forte de Gaia | September 26

In one of our Computer Vision projects, our team implemented a solution for automated surface defect detection in cold rolling production.

The objective was practical: detect defects on metal sheets in real time, reduce reliance on manual inspection, and identify issues before the material moves to the next processing stage.

The workflow is straightforward:
Metal sheet monitored → defect detected → location and type identified → operator alerted → data transferred to production systems

The system can detect and classify up to 10 types of surface defects, while automatically fixing their coordinates and defect size for further analysis and action.

Processing speed: up to 5 m/s
Detection accuracy: up to 85%
Response time: real-time

If surface inspection is still largely manual, we can assess your production line and explore how Computer Vision could support automated quality control.

itp.biz

100% zone coverage, under 2 seconds latency — here’s how we got there.

On a recent Computer Vision deployment, we built a system to track contractor working time inside a defined production zone. Timesheets don’t always show what happened on the floor, and with a large contractor workforce, that gap creates disputes during settlements.

The workflow is straightforward:
Person enters zone → role classified → presence tracked → working time recorded → available for video review

The system classifies up to 5 personnel types.
Worker coverage in monitored zone: 100%
Detection/tracking latency:

In industrial environments, safety challenges often arise in moments.
A person entering a restricted area, moving too close to hazardous equipment, or accessing a high-risk zone can create a serious safety concern before it is noticed through traditional monitoring methods.

For this project, we implemented a real-time video monitoring solution to improve visibility of personnel movement across hazardous industrial zones.
The objective was simple: help safety teams understand what is happening inside critical areas, identify risks faster, and have better information available for incident review.

The solution enabled:
✅ Real-time monitoring of personnel entering restricted zones
✅ Flexible configuration of hazardous areas based on site requirements
✅ Individual person tracking within monitored spaces
✅ Complete movement path analysis inside defined zones
✅ Recorded video review for incident investigation and safety audits

Project Outcomes:
📍 Detection accuracy: 95%+
📍 Zone management: Real-time adjustment of monitored areas
📍 Tracking capability: Individual movement monitoring
📍 Analysis: Complete movement history for review and investigation

By improving visibility across hazardous areas, the system helps safety teams better understand workplace activity and support faster decision-making when risks occur.

Have a challenging area where continuous monitoring is difficult?
Share your existing camera setup with us, and we can evaluate the possibilities.

In our Computer Vision project, the team implemented a solution to automate PPE compliance monitoring in an industrial environment.

The objective was practical: reduce dependence on manual observation and provide continuous visibility into whether required protective equipment is being used.

The workflow is straightforward:

Person detected → PPE checked → violation registered → alert sent → incident available for later review

The system recognizes 4 categories of PPE:

➡️ Helmet
➡️ Safety glasses
➡️ Respiratory protection
➡️ Firesuits

When a violation is identified, the event can be sent to the appropriate systems for further action. Historical video and analytics can also be used to review each individual case.

Detection accuracy: 95%+

For HSE teams, the value is not simply detecting a person on camera. It is automatically checking whether the required protective equipment is actually present and creating a record when it is not.

If PPE compliance is still monitored mainly through manual checks, we can assess one camera view from a work zone and evaluate whether this approach is suitable for the environment.

In one of our Computer Vision projects, our team implemented a solution to automate slag pot processing control in an industrial environment.
The goal was to improve process visibility, reduce manual monitoring efforts, and provide better tracking of slag pot conditions during each locomotive pass.

The workflow is straightforward:
Slag pot detected → Pot condition classified → Spraying process monitored → Violation recorded

The system automatically detects and tracks slag pots, identifies their condition (empty / with slag), monitors spraying activities, and captures process deviations for further review.

When a violation is detected, the system records the event with relevant information to help teams analyze and improve the process.

Key capabilities include:
• Slag pot detection and tracking
• Pot state classification
• Spraying zone and process monitoring
• Automated violation detection

Detection accuracy: 98% ✔️
For the operations team, the solution provides better visibility into slag pot processing and helps review important process conditions during each locomotive pass.

If you are looking to improve monitoring and control within your industrial processes, InTech Partner | ITP team can help explore how Computer Vision can support your operations.

A successful S/4HANA migration in pharma isn’t just about moving to a new system.

In a regulated environment, the decisions made before and during the migration can matter just as much as the technology itself.

From our experience working on pharmaceutical transformation projects, five lessons consistently stand out:

→ Design around regulated processes, not just SAP modules
→ Define validation scope early
→ Give data migration clear ownership
→ Test exceptions, not only the happy path
→ Treat user adoption as part of maintaining control

Compliance shouldn’t be something you try to prove at the end. It should influence how the migration is designed from the start.

Swipe through the carousel for the five practical lessons ➡️

Tell us the compliance risk you’re most concerned about. Our pharma team would be happy to discuss it.

GxP ITP

Your best engineers shouldn’t spend hours rebuilding routing cards.
Yet that’s still part of the daily workload for many manufacturing teams — reviewing drawings, checking old projects, repeating calculations, and turning the same engineering knowledge into new documentation.

AI can take some of that repetitive work off the team’s plate.
Technical drawings and manufacturing parameters can be used to create structured draft process plans and routing cards, with engineers reviewing and approving the results.

💬 Comment “AI Workflow” if you’d like to see how this workflow can work in practice.