Technical Demo
Demo — Video to Structured Vision Features
Proving that video can be reliably converted into structured vision data. This is a static capability showcase, not an interactive demo.
Pipeline Architecture
From video input to feature output
Input
- RTSP streams
- Direct camera
- Video files
Processing
- Object detection
- Face recognition
- Object tracking
Output
- REST API
- Webhook
- Structured events
Screenshots
Live run examples
Placeholder — to be filled with real run screenshots: video stream, detection boxes, feature output, API response.
Feature Output
Vision Feature structure
The platform outputs a unified structured vision feature. Applications consume only standardized data, independent of algorithm implementation.
// Vision Feature — unified structured output { "type": "face", "timestamp": "2026-07-23T10:42:18.412Z", "source": { "stream_id": "rtsp://camera-01/track1", "frame": 18472 }, "object": { "class": "person", "bbox": [320, 118, 488, 402], "embedding": "<512-dim feature vector>", "attributes": { "track_id": 17 } }, "confidence": 0.964, "metadata": { "model": "face-det-v2", "device": "rk3588-edge" } }
API Example
REST API call
# Subscribe to vision features from a stream curl -X POST https://api.bangvision.ai/v1/streams \ -H "Content-Type: application/json" \ -d '{ "input": "rtsp://camera-01/track1", "pipeline": ["detect", "track", "recognize"], "output": { "type": "webhook", "url": "https://your.app/hooks/vision" } }' # Response { "stream_id": "st_8f3a...", "status": "running", "features_url": "https://api.bangvision.ai/v1/streams/st_8f3a/features" }
Performance
Reference performance
Placeholder — to be filled: Latency / FPS / Device (e.g. Jetson Orin / RK3588).
Latency
End-to-end
Placeholder — to be filled.
Inference
Placeholder — to be filled.
Throughput / Device
FPS
Placeholder — to be filled.
Tested Devices
Placeholder — e.g. Jetson Orin / RK3588.
Get Started
See the platform in action
Explore the technical demo, or tell us about your use case to request a live walkthrough.