SafetyEye SE-ASI-202/ME(С) - Video Analytics Server

Face, People and Vehicle Recognition from 64 Full HD Video Streams — on One Server

When there are many cameras, an operator cannot watch every screen, and finding a particular person or car in the archive turns into manual rewinding of recordings. The SafetyEye SE-ASI-202/ME(C) takes this over: using GPU accelerators and deep learning algorithms, the server detects, recognizes and matches faces, people and vehicles on its own, builds structured metadata and stores the results. It analyzes up to 64 Full HD (1080p) video streams at once, processes up to 256 face images per second and stores up to 80 million image records with metadata. Analytics are accessed through a web interface.

👤 Who it is for

• Safe City system integrators: city and regional projects that need a central analytics node for many cameras.
• Law enforcement: checking faces and vehicles against watchlists with alerts, searching for people by photo or body image, analyzing routes and movement chronology.
• Transport hubs and educational institutions: centralized video surveillance and situational monitoring with a history of detections.
• Industrial enterprises and critical infrastructure: vehicle monitoring and automatic detection of matches with set criteria.
• Growing projects: a cluster of up to 5 servers with load distribution and automatic failover if a node fails.

⚙️ What is inside and what it means in practice

• Video streams: up to 64 Full HD (1080p) video streams at once on a single server. Streams can arrive in real time (H.264, H.265) or from recordings.
• Photos and matching: analysis of up to 256 face images per second (2 MP), matching up to 30 pairs per second. Database limits: up to 128 libraries and up to 3 million watchlist records — compare them with your project.
• Archive: up to 80 million image and metadata records — 20 million hot data + 60 million cold data. Accumulated data is used for retrospective analysis and investigations. Storage subsystem: 5 x SSD 480 GB + 5 x HDD 12 TB SATA on a PERC H755 hardware RAID controller, 12 bays for 3.5" drives.
• Hardware: Intel Xeon Silver 4516Y+, 128 GB DDR5 ECC RDIMM, 2 KT2 cards (2x4 chips). Power consumption is 290 W nominal, 319 W maximum. Two 1400 W power supplies in 1+1 redundancy with hot-swap.
• Interfaces and form factor: 6 x 1Gb/s LAN, iDRAC9 Enterprise with a dedicated iDRAC network port. Rack 2U, 715.5x482x86.8 mm, 21.9 kg. Operating temperature from +5°C to +35°C.

The server accepts real-time video streams in H.264 and H.265 with resolution from 720p to 8 MP and a frame rate of 12 to 30 fps (25 recommended). Recordings can also be imported separately (a single file up to 10 GB, up to 100 GB in total) in TS, DaHua es, SVAC, MBF, PS, MP4, AVI, ASF, HIK es and Tian Di es formats — analysis of such recordings is accelerated up to 40 times.

The number of channels depends on camera resolution: for face recognition it is 64 channels at 2 MP, 48 at 2–5 MP and 32 at 5–8 MP. How many streams the server handles when face recognition and people and vehicle analysis run at the same time, check with a manager.