KR|EN

Core Technology

Vision AI built for the factory floor

It captures human movement even in harsh industrial scenes — and understands what that movement means.

import biovis_ai as bv model = bv.PoseEstimator( mode="industrial_top_view") keypoints = model.inference(frame) # 15 keypoints

BV-POSE

In-house pose AI

A 15-keypoint engine designed for high-angle, wide-angle factory CCTV.

01

Optimized for Top-View

Tracks 15 joints in high-angle and wide-angle distortion, not just frontal views.

02

Adaptive Performance

Choose a lighter or higher-precision model to match on-site hardware.

03

Robust Tracking

Minimizes ID switching among mixed workers and equipment.

Architecture
Custom YOLOv8-based Network
Input Resolution
640px
Inference Speed
Optimized per on-site hardware (measured on deployment)
Output
Normalized Keypoints (JSON / NPY)
Keypoints
15
Multi-person tracking
Supported
Runtime
On-premises, factory PC

Behavior analysis

Beyond posture, we read behavior

Time-series pose plus each worker's movement habits. Not a fixed threshold — we read what is different from that person's usual pattern.

Factory CCTV with skeleton overlay marking lower-back strain Analysis · lower-back strain

The full pipeline

Full pipeline: detect, pose, behavior analysis, strain assessment, dashboard

Vision AI, tailored to your site.

We develop custom video-analysis software for unusual camera angles and environments.

Talk about a build