dadcam

Automatic camera trap media processor for the Steam Deck.

Insert a compact flash card from a trail/wildlife camera and dadcam will:

1. Detect the card via udev
2. Run YOLOv8 animal/person detection on every image and video clip
3. Sort files into detections/ or no_detections/ on your destination drive
4. Remove originals from the CF card after verified copy
5. Write a Markdown report summarising the run

Everything runs locally on the Deck's CPU — no cloud, no internet required after setup.

---

Requirements

• Steam Deck (SteamOS 3.x) in Desktop Mode
• The CF card reader you use (USB or dock)
• ~2 GB free space for the Python virtualenv + YOLOv8 model weights

---

Installation

1. Get the code onto the Deck

# Option A — git
git clone https://github.com/you/dadcam.git ~/dadcam

# Option B — copy from another machine
scp -r /path/to/dadcam deck@steamdeck.local:~/dadcam

2. Run the installer

Run as the deck user (not root):

cd ~/dadcam
bash install.sh

This will:

• Create ~/dadcam/venv/ with an isolated Python environment
• Install CPU-only PyTorch (~200 MB) and all other dependencies
• Write a default config to ~/.config/dadcam/dadcam.conf
• Install ~/.config/systemd/user/dadcam@.service (no root required)
• Enable systemd linger via sudo loginctl enable-linger deck so the service activates at boot without a desktop login (does not disable the read-only root fs)

3. Edit your config

nano ~/.config/dadcam/dadcam.conf

At minimum, set the destination — typically your SD card:

[paths]
destination = "/run/media/deck/MY_SDCARD/dadcam_output"

4. Whitelist your CF card reader and install the udev rule

Insert any CF card into your reader, then run the setup wizard once:

sudo ~/dadcam/venv/bin/python ~/dadcam/dadcam.py --setup

The wizard will:

• Detect the inserted card and display the reader's serial and device info
• Ask you to confirm whitelisting the reader device
• Write /etc/udev/rules.d/99-dadcam.rules (briefly disables SteamOS read-only fs)

From this point, inserting any CF card into the whitelisted reader triggers processing automatically.

---

Usage

Automatic (normal use)

Insert any CF card into the whitelisted reader. Processing starts within a few seconds. Watch progress in real time:

journalctl --user -u 'dadcam@*.service' -f

Manual — test on a folder

~/dadcam/venv/bin/python ~/dadcam/dadcam.py --source /path/to/photos

Dry run — preview without moving files

Pass --dry-run with any --process or --source invocation to run the full detection pipeline and write a report, but without copying or deleting anything:

# Dry run on a directory
~/dadcam/venv/bin/python ~/dadcam/dadcam.py --source /path/to/photos --dry-run

# Dry run on a device
~/dadcam/venv/bin/python ~/dadcam/dadcam.py --process --device /dev/sda1 --dry-run

Each file will be logged as DRY-RUN would move … and the report will indicate that no files were moved or removed.

Manual — process a specific device

~/dadcam/venv/bin/python ~/dadcam/dadcam.py --process --device /dev/sda1

View recent reports

~/dadcam/venv/bin/python ~/dadcam/dadcam.py --report --last 5

List whitelisted drives

~/dadcam/venv/bin/python ~/dadcam/dadcam.py --list-whitelist

Add another CF card reader

Re-run setup with a card inserted in the new reader:

sudo ~/dadcam/venv/bin/python ~/dadcam/dadcam.py --setup

Once the reader is whitelisted, any card you insert into it will be processed automatically.

---

Output structure

<destination>/
├── detections/          # files where an animal or person was detected
│   └── IMG_0042.jpg
├── no_detections/       # files with nothing of interest
│   └── IMG_0001.jpg
└── reports/
    └── 2026-02-18_14-30-00.md

Original files are removed from the CF card only after the copy is SHA-256 verified. Files already present at the destination with matching content are skipped (deduplication).

---

Detection

Uses YOLOv8n (Ultralytics) running on the Deck's CPU.
Model weights (~6 MB) are downloaded automatically on the first run to ~/.local/share/dadcam/models/.

Default classes detected: person, bird, cat, dog, horse, sheep, cow, elephant, bear, zebra, giraffe.

Customise in ~/.config/dadcam/dadcam.conf:

[detection]
model = "yolov8n"
confidence_threshold = 0.35
classes_of_interest = ["person", "bird", "deer", "bear"]

For video files, frames are sampled every 30 frames (configurable via frame_sample_interval). A video is marked as a detection if any sampled frame contains a match.

---

Configuration reference

~/.config/dadcam/dadcam.conf (TOML format):

[paths]
destination = "/home/deck/Pictures/dadcam_output"

[detection]
model = "yolov8n"               # yolov8n | yolov5s
confidence_threshold = 0.35
classes_of_interest = [
  "person", "bird", "cat", "dog", "horse", "sheep",
  "cow", "elephant", "bear", "zebra", "giraffe"
]

[video]
frame_sample_interval = 30      # check every N frames

[report]
keep_reports = 50               # oldest reports pruned automatically

[logging]
level = "INFO"                  # DEBUG | INFO | WARNING | ERROR
log_file = "/home/deck/.local/share/dadcam/logs/dadcam.log"

---

File layout

Path	Purpose
~/dadcam/	Scripts and virtualenv
~/.config/dadcam/dadcam.conf	User configuration
~/.config/dadcam/whitelist.conf	Whitelisted CF card reader serials (device-level)
~/.config/systemd/user/dadcam@.service	systemd user service unit
~/.local/share/dadcam/models/yolov8n.pt	Cached model weights
~/.local/share/dadcam/logs/dadcam.log	Persistent log file
/etc/udev/rules.d/99-dadcam.rules	udev device trigger ⚠️ see note below

SteamOS update note: Major SteamOS updates reset /etc/udev/rules.d/. After an update, re-run sudo ~/dadcam/venv/bin/python ~/dadcam/dadcam.py --setup to restore the udev rule. Everything else (config, whitelist, service, models) lives in your home directory and is unaffected.

---

Troubleshooting

Card inserted but nothing happens

# Check the udev rule is installed
cat /etc/udev/rules.d/99-dadcam.rules

# Check the systemd service exists
ls ~/.config/systemd/user/dadcam@.service

# Check linger is enabled
loginctl show-user deck | grep Linger

# Re-run setup if any of the above are missing
sudo ~/dadcam/venv/bin/python ~/dadcam/dadcam.py --setup

"Device not whitelisted" in logs

The card's reader serial (or UUID for legacy entries) doesn't match the whitelist. Run setup again with a card inserted in the reader.

~/dadcam/venv/bin/python ~/dadcam/dadcam.py --list-whitelist

Model download fails

Ensure the Deck has internet access on first run, or manually download yolov8n.pt from ultralytics releases and place it in ~/.local/share/dadcam/models/yolov8n.pt.

Check recent run logs

journalctl --user -u 'dadcam@*.service' --since "1 hour ago"

---

Project layout

dadcam/
├── dadcam.py        # CLI entry point
├── scanner.py       # walk source directory, enumerate media files
├── detection.py     # YOLOv8 inference on images and video
├── sorter.py        # copy + SHA-256 verify + remove source
├── reporter.py      # write Markdown run report
├── setup_mode.py    # interactive udev + systemd setup wizard
├── config.py        # TOML config loader
├── whitelist.py     # whitelist read/write
├── requirements.txt
├── install.sh       # bootstrap virtualenv and dependencies
├── README.md
└── SDD.md           # software design document

---

License

MIT