Warehouse rfid controller

Warehouse Access RFID + Computer Vision

Graduation project prototype for warehouse access control and employee monitoring using ESP32 RFID, cameras, YOLO detection, tracking, Re-ID, optional face recognition, fusion logic, SQLite, and a Flask dashboard.

Main Components

• firmware/: ESP32 Arduino firmware for RC522 RFID, LCD I2C, buzzer, servo gate, WiFi, NTP, and MQTT event publishing.
• python_cv/: main PC-side application for camera processing, Re-ID, InsightFace hybrid recognition, RFID/CV fusion, SQLite storage, and dashboard.
• hardware/: KiCad PCB source files, custom footprints, 3D models, and Gerber export for the RFID controller board.
• docs/: submission notes, demo guide, diagrams, and report assets.
• experiments/, python_cv/tools/, python_cv/tests/: debug, calibration, data collection, and validation utilities.

The active PC pipeline is python_cv/main.py. The nested folder warehouse-access-rfid-cv/ is an older/stale copy and should not be used for the final submission without review.

Setup

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
pip install -r requirements.txt

If CUDA-specific PyTorch wheels are required, install PyTorch first using the official command for the target GPU/CUDA version, then install the remaining requirements.

Copy the clean runtime example and edit local values:

Copy-Item .env.example python_cv\.env

Do not commit python_cv/.env, RTSP passwords, WiFi passwords, raw videos, face/person galleries, or private datasets.

Dataset and Local Data Path

AI datasets, raw videos, review crops, galleries, and model artifacts are local runtime assets. They are intentionally not part of the source-code submission.

Default local dataset root:

D:\warehouse_dataset

Default subfolders used by tools:

D:\warehouse_dataset\dataset_crops
D:\warehouse_dataset\raw_videos_full
D:\warehouse_dataset\review_crops_reid_yoloseg
D:\warehouse_dataset\raw_videos_from_imou_sd

These defaults are configured through .env:

WAREHOUSE_DATASET_ROOT=D:/warehouse_dataset
DATASET_CROPS_ROOT=D:/warehouse_dataset/dataset_crops
RAW_VIDEO_ROOT=D:/warehouse_dataset/raw_videos_full
REVIEW_CROP_ROOT=D:/warehouse_dataset/review_crops_reid_yoloseg
IMOU_SD_VIDEO_ROOT=D:/warehouse_dataset/raw_videos_from_imou_sd

If another machine uses a different drive or folder, edit python_cv/.env instead of changing code. To copy local data from the old external F: drive to D:, review and run:

.\scripts\migrate_warehouse_dataset_F_to_D.ps1
.\scripts\verify_dataset_paths.ps1

The migration script copies data only; it does not delete anything from F:.

Run Demo

From the repository root:

cd python_cv
python main.py

Default demo mode can run CV-only with Zone 2 webcam and InsightFace enabled. For RFID + CV end-to-end, set ENABLE_MQTT=true, configure the MQTT broker, flash the ESP32 firmware, and start the broker before running python main.py. Create firmware/config/device_config.h from firmware/config/device_config.example.h before flashing the ESP32.

Dashboard:

http://localhost:5000

Detailed demo steps are in docs/RUN_DEMO.md.

Submission Notes

Before submitting, review:

• CLEANUP_AUDIT_REPORT.md
• SUBMISSION_STRUCTURE_PROPOSAL.md
• docs/CLEAN_SOURCE_CHECKLIST.md
• CLEANUP_COMMANDS_DRAFT.ps1

The current repo contains local debug outputs, model files, gallery pickles, raw videos, and real credentials in firmware. Clean or archive them before handing the project to a reviewer.