How pulkitpareek18/aeris works
AERIS is a camera-free Wi-Fi CSI sensing research system. Its principal V1 workflow turns public or recorder-captured CSI into prepared model inputs, runs presence inference on a live Raspberry Pi stream, and sends telemetry to a dashboard. The graph also preserves the recorder and dataset-conversion capabilities. Edges are limited to README-documented flows and relationships visible in the supplied excerpts or SOURCE INDEX; dashed edges mark unverified integration details.
Main entry points: live_server.py Live inference server · schema.py CSI record schema · aeris_recorder_server.py Recorder service · build_corpus.py Corpus builder
Components 21
Capture and recording
- Recorder service
aeris-recorder/aeris_recorder_server.py - Session metadata
aeris-recorder/aeris_db.py
Dataset ingestion
- Presence loader
ml/src/aeris_ml/datasets/presence_movement.py - Recorder PCAP loader
ml/src/aeris_ml/datasets/aeris_pcap.py - Dataset adapters
ml/src/aeris_ml/datasets - Dataset registry
ml/src/aeris_ml/datasets/registry.py
ML preparation and inference
- CSI record schema
ml/src/aeris_ml/schema.py - Label mapping
ml/src/aeris_ml/labels.py - Record validation
ml/src/aeris_ml/validation.py - Signal preparation
ml/src/aeris_ml/preprocessing.py - Corpus builder
ml/src/aeris_ml/build_corpus.py - Model training
ml/src/aeris_ml/train.py - Perception model
ml/src/aeris_ml/models.py - Room calibration
ml/src/aeris_ml/calibrate.py - Nexmon stream decoder
ml/src/aeris_ml/nexmon_stream.py - Pi geometry adapter
ml/src/aeris_ml/pi_adapter.py - Live inference server
ml/src/aeris_ml/live_server.py
Live presentation
- Live dashboard
dashboard/src/app/page.tsx
Other components
- Researcher
- Raspberry Pi
- Public CSI data
Connections 24
- Researcher → Recorder service: records sessions
- Recorder service → Session metadata: stores metadata
aeris-recorder/aeris_recorder_server.py - Raspberry Pi → Recorder service: provides capture
README.md - Public CSI data → Presence loader: loads recordings
ml/src/aeris_ml/datasets/presence_movement.py - Recorder service → Recorder PCAP loader: exports sessions
ml/src/aeris_ml/datasets/aeris_pcap.py - Public CSI data → Dataset adapters: loads datasets
- Dataset registry → Dataset adapters: selects adapters
ml/src/aeris_ml/datasets/__init__.py - Presence loader → CSI record schema: creates records
ml/src/aeris_ml/datasets/presence_movement.py - Recorder PCAP loader → CSI record schema: creates records
ml/src/aeris_ml/datasets/aeris_pcap.py - Dataset adapters → CSI record schema: creates records
ml/src/aeris_ml/datasets/aeris_pcap.py - CSI record schema → Label mapping: uses vocabulary
ml/src/aeris_ml/schema.py - Corpus builder → Signal preparation: prepares signals
ml/src/aeris_ml/build_corpus.py - Corpus builder → Presence loader: loads corpus data
- Corpus builder → CSI record schema: serializes records
- Signal preparation → Record validation: validates records
ml/src/aeris_ml/validation.py - Model training → Perception model: trains network
ml/src/aeris_ml/train.py - Researcher → Live inference server: opens live session
README.md - Live inference server → Raspberry Pi: streams CSI
README.md - Live inference server → Nexmon stream decoder: decodes packets
ml/src/aeris_ml/live_server.py - Live inference server → Pi geometry adapter: adapts geometry
ml/src/aeris_ml/live_server.py - Live inference server → Perception model: runs inference
ml/src/aeris_ml/live_server.py - Live inference server → Live dashboard: sends WebSocket data
README.md - Researcher → Live dashboard: views telemetry
README.md - Room calibration → Perception model: calibrates checkpoint
ml/src/aeris_ml/calibrate.py