terminal://developer-profile
Vlad Voropaev
Computer Vision Engineer
I build multi-camera video analytics systems from dataset design and held-out evaluation through RTSP integration, tracking, pose logic, field iteration, and event evidence.
$
analyze rtsp --detect --track --reason▌
mode: developer-profile · build: industrial video analytics
help · projects · research · stack · contact · matrix · clear
ready: explore the portfolio or run a command
Object detection
Object tracking
Pose estimation
PPE & zone analytics
OCR
6
camera views in a crane operator safety system
17,696
CCTV frames reviewed for a multi-camera pilot
0.920
held-out mAP@0.50 for the crane safety detector
01://INDUSTRIAL-CV
Computer Vision systems built for real operating constraints
Selected industrial Computer Vision systems developed for safety, monitoring, and multi-camera video analytics
PAID CUSTOMIZATION
TYPE://SAFETY-VISION
Crane Operator Safety
A six-camera system developed end to end, from dataset design and held-out evaluation to tracking, pose logic, field debugging, and saved event evidence
- 1,163 images and 7,499 labeled objects; YOLOv8m detector
- 0.932 precision, 0.878 recall, 0.920 mAP@0.50, and 0.707 mAP@0.50:0.95
- Active-view selection, ByteTrack identities, wrist-to-pendant pose association, helmet checks, and temporal voting
- A white-helmet field failure drove targeted fine-tuning and revised weights
Python · PyTorch · Ultralytics YOLO · ByteTrack · pose estimation · event media
INDUSTRIAL PILOTS
TYPE://VIDEO-ANALYTICS
RTSP Safety Analytics
Modular video analytics for PPE, danger zones, tracking and counting, equipment use, worker activity, and overlapping camera views
- Offline PPE validation: 0.906 mAP@0.50 and 0.772 mAP@0.50:0.95 on 579 images / 1,748 objects
- 17,696 CCTV frames reviewed from eight feeds; 1,156 relevant frames curated
- View-specific zones reduced long-range and duplicate detections across overlapping views
- Low-resolution worker activity used tracking, pose, fixed zones, temporal stabilization, reason labels, and event evidence
RTSP · OpenCV · detection · tracking · PPE · polygon zones · FastAPI · Docker · PostgreSQL
02://RESEARCH
NeuroQuest: video understanding to generated comic
An implemented research system with a traceable path from live video to perception, narrative, generated imagery, and PDF output
IEEE ACDSA 2024FIRST AUTHOR
Automatic generation of neurocomics based on computer vision system data
Built the end-to-end RTSP-to-comic pipeline combining detection, tracking, face and action recognition, scene interpretation, language and image generation, and PDF assembly. Led the paper writing and presented the work orally at ACDSA 2024
video
perception
narrative
generated visuals
IEEE Xplore 10467698
03://ENGINEERING-LENS
From camera stream to evaluated system
The strongest work spans the model, the runtime logic around it, and the evidence needed to improve it in the field
A
Visual perception
Object detection, tracking, pose estimation, OCR, image preprocessing, PPE and zone analytics
B
Video intelligence
RTSP ingestion, multi-camera logic, temporal stabilization, event reasoning, evidence capture, and activity recognition
C
Evaluation and iteration
Dataset design, annotation, held-out evaluation, per-class error analysis, field failure review, and targeted retraining
D
System integration
FastAPI services, Docker, PostgreSQL-backed event flows, GPU inference, and engineering documentation
Python
PyTorch
OpenCV
Ultralytics YOLO
ByteTrack
Deep SORT
RTSP
FastAPI
Docker
PostgreSQL
04://PUBLIC-CODE
Open-source engineering and earlier work
Open-source product engineering followed by selected earlier repositories