Wesense has always been my favorite. A very good session on the product, architecture
- AI through IoT
- Situational Awareness
- Smart Cameras, Self Driving, Cars
- Videos - Sequences of frames
- Detections, tracking through frame for Vision Apps
Vision Apps
- People Counting
- Features + Face Registered
- Customer Attribute Tracking (Glasses)
- In Class Emotions Tracking / Participation score
- Dispensed Cups
- Key Dispensing kit
- Key - Car Mapping - OCR + Extraction
Production Stack
- Remote SSH Provider
- Onboard GPU
- Tracking Area Identified for each camera
- Meta data inside a region
- Notification Systems
- Camera Positioning
- Multiple camera sync
- Object reidentification with redis
- Occlusion - Multiple cameras
- Intel openvino tensorflow
- Heavier inference on cloud
- Lighter inference on edge device
- Age and Gender model on cloud
- Focal Length of Camera, Height and other customer identifiers to match
- Batch processing on edge (patented)
- Preprocessing to determine which one is processed
- Determine which frames to be picked up
- Not running all models on all time
- CCTV much better than drones on few cases outdoor monitoring
- Team to label data
- 1.5 years to reach this level of accuracy
Happy Learning!!!
No comments:
Post a Comment