ClickMasters builds video analytics systems for B2B companies across the USA, Europe, Canada, and Australia. Object detection and tracking in live video streams detect, identify, and follow every object of interest. Retail analytics customer flow, dwell time, queue length, conversion zone heatmaps. Crowd analysis and density estimation. Action recognition detect specific events and behaviours. License plate recognition. All analytics delivered as real-time event streams your operations team can act on.
Years Experience
Projects Delivered
Client Satisfaction
Support Available
DeepSORT (Deep Simple Online and Realtime Tracking) is the production standard for object tracking in video analytics. It extends SORT by adding a deep appearance descriptor (a CNN embedding of the object's appearance) that enables re-identification when objects are occluded or cross. Tracking workflow: detection (YOLO v8 produces bounding boxes each frame), Kalman filter (predicts object location in next frame), Hungarian algorithm (matches detections to existing tracks), appearance feature extraction (CNN embedding of the detected object, used for matching after occlusion), and track management (initialise new tracks, terminate tracks that exit the frame). Output: each tracked object receives a persistent ID across frames, enabling trajectory analysis (path, dwell time, entry/exit zones). ClickMasters uses DeepSORT as the default tracker for all video analytics applications.
Privacy-preserving video analytics is a design principle, not an afterthought. For retail and workplace analytics, ClickMasters systems output aggregate metrics and anonymised trajectories not facial recognition or persistent individual identification. Technically: object tracking assigns temporary IDs per session (ID does not persist across camera views or days), bounding box blurring (faces and identifying features blurred in any stored video clips), no biometric data collection, aggregate dashboard outputs (counts, heatmaps, density maps no individual-level data), and data retention policies (raw video not stored beyond the minimum required buffer; derived analytics data retained per policy). GDPR compliance: processing is based on legitimate interest for operational safety and efficiency, with appropriate signage notifying visitors. ClickMasters documents the data processing activities for GDPR Article 30 records.
What we deliver
05 capabilities
ClickMasters operates as a full-stack video analytics partner — product strategy, UI/UX, engineering, cloud infrastructure, QA, and ongoing support in one delivery model.
YOLO v8 (detection) + DeepSORT or ByteTrack (tracking) detect objects each frame, assign persistent IDs, track trajectory across frames. Output: per-object trajectory (entry time, path, exit time, dwell zone), object count, object classification. Use cases: retail customer tracking (anonymous), vehicle tracking on logistics sites, equipment tracking. Real-time or batch processing. Privacy-preserving: trajectories without storing identifiable images.
Detect specific human or machine behaviours from video sequences. Approaches: two-stream networks (spatial + optical flow), SlowFast networks (Facebook dual temporal pathway), Video Swin Transformer. Use cases: safety compliance (PPE detection hard hat/high-vis worn?), workplace safety event detection (slip/fall), retail customer behaviour (picking up product, comparing items), manufacturing process monitoring (correct assembly sequence).
In-store analytics from existing CCTV cameras: footfall counting (accurate people counting at entrance/exit), customer flow heatmaps (where customers spend most time, which aisles are undervisited), queue length estimation (number waiting at checkout trigger staffing alerts), zone dwell time, conversion zone analysis (percentage who enter product zone that convert to checkout). Dashboard: real-time and historical analytics, shift comparison, A/B zone experiment analysis.
Automatic Number Plate Recognition for parking, access control, and logistics. Pipeline: vehicle detection (YOLO v8), plate localisation, character recognition (YOLO-based character segmentation + OCR, or end-to-end CNN), database lookup (check plate against access whitelist, trigger gate open/close). Accuracy: 95-99% on clean, well-lit footage. Multiple plate standards (UK, US, EU). Real-time: <200ms end-to-end.
Density estimation and crowd monitoring: CSRNet or MCNN (crowd counting estimate density without tracking individuals), occupancy monitoring (real-time room/zone capacity compliance alert when exceeds threshold), crowd flow analysis (direction and speed of movement event safety planning), social distancing monitoring (historical analysis only inter-person distances). Privacy-preserving: aggregate counts and density maps, not individual tracking.
Why choose us
05 advantages
We combine architecture discipline, transparent delivery, and long-term partnership — so your investment translates into measurable business results, not just shipped code.
Persistent object IDs across frames entry, path, exit, dwell time analysis | Basic: Frame-by-frame detection only (no object continuity)
SlowFast + Video Swin Transformer detect behaviours over time, not just objects | Basic: Object detection only (what, not what action)
RTSP stream ingestion + TensorRT optimisation (3-5x faster) + Kafka event streaming | Basic: Batch processing on recorded video only
Edge deployment (Jetson Orin) low latency, no cloud round-trip | Basic: Cloud-only (network latency, video leaves site)
Anonymous trajectories, aggregate density maps, no facial identification | Basic: Stores identifiable images
500+
Companies served
4.9/5
Client rating
15+
Years in delivery
A clear, structured approach — so you always know what happens next and why.
Phase 1
Camera inventory (RTSP endpoints, resolution, placement), use case analysis (tracking, action, counting, crowd), latency requirements (real-time <100ms vs batch), privacy constraints, infrastructure plan. Deliverable: Video Analytics Architecture.
Phase 2
YOLO v8 fine-tuning on custom object classes (if needed), DeepSORT tracking integration, RTSP stream ingestion (GStreamer/FFmpeg), trajectory output, dashboard (heatmaps, counts). Deliverable: Detection + Tracking Pipeline.
Phase 3
SlowFast or Video Swin Transformer training on custom action dataset (if labelled video available), action detection pipeline, real-time alerting. Deliverable: Action Recognition System.
Phase 4
Footfall counting (entrance/exit zones), heatmap generation, queue length estimation, CSRNet for crowd density, dashboard (real-time metrics, historical trends). Deliverable: Analytics Dashboard.
Modern tools we use to build scalable, secure applications.
Languages & Frameworks
Data Processing
Infrastructure
Deep expertise across various sectors with tailored solutions
Pricing
Transparent pricing tailored to your business needs
* All prices are estimates and may vary based on requirements.
To build scalable, intelligent custom software development solutions that empower businesses to grow, automate, and transform in a digital-first world.
We are not building software. We are architecting the infrastructure of tomorrow—systems that think, adapt, and grow alongside the businesses they power.
Amjad Khan
Chief Executive Officer
12+
Years Exp
300+
Success
98%
Retention
Common Inquiries
Can't find the answer you're looking for? Please chat to our friendly team.
Discover how we can help transform your business through our comprehensive services, real-world case studies, or our full solutions portfolio.