Search Your CCTV Footage
Like a Search Engine

Banglai Cloud CCTV Analytics is an AI-powered CCTV video analysis platform built in Bangladesh. It lets you upload hours of CCTV footage (up to 2GB per file) and search through it like a search engine. The system uses a 7-stage hybrid pipeline — YOLOv11s for object detection, ByteTrack for tracking, YOLO-World for open-vocabulary detection, and OpenCLIP for semantic search. You can search by typing natural language ("red car", "person in blue shirt"), uploading a reference image, or finding similar detections. 24 hours of footage is processed in approximately 1 hour. Pricing starts at ৳0/month (free) with paid plans at ৳490/month (Individual) and ৳990/month (Commercial).

The system processes your video by sampling 1 frame every 3 seconds, running YOLOv11s object detection on each sampled frame, tracking objects across frames with ByteTrack, and generating a 512-dimensional OpenCLIP vector embedding for each detected object crop. These embeddings are stored in PostgreSQL with pgvector. When you search by text, your query is encoded into the same vector space and the system finds the closest matching detections. When you search by image, the reference image is encoded and matched. This means you can search for concepts like "person in red shirt" without manually tagging anything — the AI understands visual semantics.

24 hours of CCTV footage is processed in approximately 1 hour (24x real-time speed). This is achieved through smart frame sampling — the system processes 1 frame every 3 seconds instead of every frame. A person or vehicle is typically visible for 5-30 seconds, so 3-second sampling catches them 1-10 times. Additionally, MOG2 motion gating skips frames with no activity (nighttime, empty rooms), which can reduce processing time by 50-80% for typical CCTV footage. Progress is shown in real-time with ETA.

The platform supports MP4, AVI, MOV, MKV, WebM, FLV, WMV, and M4V formats. The maximum file size is 2GB, which can hold 12+ hours of CCTV footage depending on resolution and bitrate. Files are uploaded directly to the backend via Nginx with streaming (no buffering), so even large files upload smoothly.

Yes. The platform supports direct RTSP camera ingestion on the VPS. You configure your camera RTSP URL in the dashboard, and a background worker connects to the camera, reads frames, runs YOLOv11s detection and ByteTrack tracking, and publishes live detections to your dashboard via WebSocket. Reconnection with exponential backoff is built in. Privacy mode and cooldowns can be configured per camera.

Yes. Every detection crop can be downloaded individually as a WebP image with a descriptive filename. You can also download all search results as a ZIP file containing all crop images plus a CSV manifest with metadata (object class, camera ID, confidence, timestamp, trajectory ID). Downloads are available on all plans including the free tier.

The YOLOv11s model detects people, cars, trucks, buses, motorcycles, bicycles, dogs, and cats by default. The YOLO-World open-vocabulary model can detect custom objects specified by text — including bags, backpacks, suitcases, handbags, knives, and any other object class you define. This means you can search for "person with backpack" or "truck" and the system will find matching detections.

Pricing is in Bangladeshi Taka (BDT). Free plan: ৳0/month — 2 cameras, 10 uploads/month, 5,000 detections, 7-day retention. Individual plan: ৳490/month — 3 cameras, 10 uploads/month, 25,000 detections, 30-day retention. Commercial plan: ৳990/month — 10 cameras, 50 uploads/month, 250,000 detections, 90-day retention, plus case export and analytics. Custom software plans are available via WhatsApp at +8801610215271 for unlimited cameras, custom detection models, and on-premise deployment.

Yes. Banglai Cloud is built in Bangladesh by ATAULLAH Digital. The platform is hosted on a VPS and accessible worldwide at banglai.cloud. Pricing is in Bangladeshi Taka (BDT). Payment is via bKash. Support is available in English and Bengali. WhatsApp support: +8801610215271.

Traditional CCTV monitoring requires a human to watch footage in real-time or scrub through recordings manually. CCTV analytics uses AI to automatically detect and index every person, vehicle, and object in the footage, then lets you search through all detections instantly by text, image, or similarity. Instead of watching 24 hours of video, you search "red car" and see every matching detection in seconds. This transforms CCTV from passive recording to an active, searchable database.