LLaVA-e Edge AI Demo

Contextual Security Alarm Intelligence

VLM-powered camera analysis that resolves false alarms automatically and escalates real threats — reducing alert fatigue in commercial facilities.

LLaVA 7B Q4_0 llama.cpp CUDA NVIDIA A10 GPU Inference optimized Jetson Orin Nano target

What is LLaVA-e?

LLaVA 7B with inference optimization for edge deployment.

num_ctx
4096
↓
512
num_predict
512
↓
150
temperature
0.7
↓
0.1
num_batch
512
↓
128
num_thread
8
↓
4

Security camera analysis

Select a scenario or upload your own image

— or upload your own —
📷

Click to upload an image

Preview

AI decision output

🔍

Select a scenario or upload an image to begin analysis

LLaVA-e analyzing image. First inference after idle: ~180s (model loading into GPU). Subsequent: ~2s...
Edge-optimized inference · NVIDIA A10 GPU
Decision
Threat level
-
Confidence
-
Inference
-
Activity detected
Reasoning
Recommended action
Running LLaVA-e · LLaVA 7B Q4_0 · inference optimized · NVIDIA A10 GPU