- Case study ↓
- AI/ML
- Computer Vision
Production Counting That Checks Every T-Shirt, Automatically
Computer vision that detects, tracks and counts T-shirts during printing, so production counts are consistent and each shirt is counted once.
The demo
See T-Shirt Counting Matching at work
- Industry
- Computer Vision
- Input
- Printing-line video
- Output
- Shirt counts, Annotated video
- Built with
- YOLO · SORT · Python · OpenCV
The problem
Manual counts can't keep up with high-volume printing
Counting T-shirts by hand during printing is slow and prone to human error, especially at high volume. Shirts move quickly and several can be in view at once, so it's easy to miss one or count it twice.
When counts are wrong, production records drift from what was actually printed. An automatic count gives an independent number that matches what came off the line.
The solution
What T-Shirt Counting Matching does
T-shirt detectionEvery shirt in the frame is found and boxed
Multi-object trackingEach shirt keeps one ID across consecutive frames
Duplicate-free countingCounts are based on tracking IDs, not individual frames
Production monitoringA running count of shirts printed, without someone tallying by hand
Modular pipelineDetection, tracking and counting run as separate stages, easy to extend
Technology
How it’s built
Every step, from the printing-line video to the results.
Step 1
Input
- Line camera
- Recorded video
Step 2
Detection
- YOLO
- Deep Learning
Step 3
Tracking
- SORT
- Python
- OpenCV
Step 4
Output
- Shirt counts
- Annotated video
Who it’s for
- Garment printersWho need accurate counts of every shirt printed per run.
- Apparel manufacturersChecking output per batch at the end of production.
- Factory automation teamsAdding vision-based counting to existing lines.
Case study
T-Shirt Counting Matching: the full case study
- Overview
An AI-powered computer vision system that automates T-shirt detection and counting during the printing process. A deep learning object detection model finds the shirts and a tracking algorithm follows each one across video frames. Built with Python and OpenCV, it reduces duplicate counting and enables automated production monitoring.
- Problem
Manual T-shirt counting during printing is time-consuming and prone to human error, especially in high-volume production. Fast movement and several shirts in the camera frame make accurate counting hard, so an automated system was needed to detect, track and count individual shirts while minimising duplicate counts.
- Solution
ML Bench developed a computer vision pipeline that uses an object detection model to identify T-shirts in video streams. Each shirt gets a bounding box that marks its position in the camera frame.
A tracking algorithm follows individual shirts across consecutive frames with unique tracking IDs. Counting by these IDs keeps duplicate counts low and enables automated monitoring of printing operations, using Python and OpenCV.
- Challenges & solutions
Individual T-shirts had to be found consistently across video frames.Integrated a YOLO-based object detection model to locate shirts and produce bounding boxes for tracking.
The same shirt appears in many consecutive frames, so frame-by-frame counting is unreliable.Implemented the SORT tracking algorithm to link detections across frames and keep an ID for each shirt.
Counting every frame separately would inflate production counts.Used tracking IDs to tell shirts already counted apart from newly detected ones.
- Results
- Automated T-shirt detection and counting
- Less dependence on manual counting
- A structured way to verify production counts
- Modular detection, tracking and counting stages, ready for future enhancements
Have Questions?
Find the right solution
for now
Yes. The same detection and tracking approach works for other garments and items on a line. The model is trained on examples of your items, so we check a sample video first.
Why ML Bench
Why teams choose ML Bench
Built for real-world footageWe build computer vision for real cameras and scans: retail, sports, healthcare, logistics and safety.
Starts with the data you haveExisting cameras, recordings or image archives are usually enough for a first proof of concept.
Custom to your setupWe adapt the model to your camera angles, image quality, classes and workflow, not the other way round.
One team, pilot to productionData, model training, dashboard, API and deployment handled by the same team.
Next AI/ML project →
Padel Sports Analytics
Built an AI platform for padel that tracks players, classifies forehand and backhand shots, and maps movement heatmaps from match video.
Free assessment
See Your Line Counted by AI
Send us a short video from your production line and we'll show you what the model picks up, free.
- UI/UX Design
- Web Development
- Mobile Apps
- SEO
- AI/ML Solutions
