Datasets & downloads

Everything the labs need, pre-bundled. Pick whichever zip matches what you're about to do; you can come back and grab the others later.

The dataset zips are too large to live in the website's Git repository, so they are hosted as a GitHub release (tag datasets). The download links below pull each zip straight from that release — no GitHub account needed.

Course-wide teaching sample

File	Size	Contents	Used in
aiecol_data_sample.zip full bundle	395 MB	200 Oregon Critters camera-trap JPGs + 97 iNaturalist bird recordings + tidy CSVs.	All labs (1–7, 11)
aiecol_cameratraps.zip images only	297 MB	200 JPGs (~3840×2880) across 7 species + bounding-box CSV (`oregon_critters_subset.csv`).	Labs 1–5, 8, 11
aiecol_audio.zip audio only	98 MB	97 short recordings (~5–30 s each) across 10 PNW species + index CSV with species, iNat URL, license.	Labs 6, 7, 11

What's inside

Camera traps — Oregon Critters subset

200 JPGs drawn from the LILA Oregon Critters training split, with one row per bounding-box annotation:

species	count
black-tailed deer	40
roosevelt elk	30
douglas squirrel	30
townsend's chipmunk	30
black bear	25
bobcat	25
empty (no animal)	20

oregon_critters_subset.csv columns: image, label, category_id, datetime, location, width, height, bbox_x, bbox_y, bbox_w, bbox_h, split, orig_id.
License: same as Oregon Critters (CC-BY).

Bioacoustics — iNaturalist Pacific NW birds

97 short recordings from 10 PNW bird species, all CC-licensed (cc-by / cc-by-sa / cc-by-nc / cc0) and downloaded from iNaturalist:

species	scientific	count
barred owl	Strix varia	10
northern spotted owl	Strix occidentalis	10
varied thrush	Ixoreus naevius	9
pacific wren	Troglodytes pacificus	10
hermit thrush	Catharus guttatus	10
pileated woodpecker	Dryocopus pileatus	10
olive-sided flycatcher	Contopus cooperi	10
swainson's thrush	Catharus ustulatus	10
dark-eyed junco	Junco hyemalis	10
american robin	Turdus migratorius	8

audio_index.csv columns: filename, species, species_latin, inat_obs_id, inat_sound_id, license, place, observed_on, obs_url.
Format: .mp3 or .wav, mono, 22.05 kHz.

Want more data? Both source datasets are much larger than what we ship. To scale up:

Oregon Critters on LILA — 99,909 images, full COCO annotations, free via AzCopy.
iNaturalist API — millions of CC-licensed observations with sounds. The course data-build scripts are shipped in the instructor/scripts/ folder if you want to grab a bigger sample.

Where to put the data

Unzip into the project root of whatever Claude Code or RStudio folder you've been working in. The labs all assume the layout below (the zip's top-level folder is called stage/; rename it to data_sample/):

your_project/
├── CLAUDE.md
├── data_sample/
│   ├── cameratraps/        200 JPGs
│   ├── oregon_critters_subset.csv
│   ├── audio/              97 mp3/wav across 10 species folders
│   └── audio_index.csv
└── ... your code ...

On macOS/Linux:

unzip aiecol_data_sample.zip
mv stage data_sample

On Windows: right-click the zip → Extract All, then rename stage to data_sample.

Checksums (for the cautious)

If you want to verify your download finished cleanly:

aiecol_data_sample.zip   MD5  f012156d9372be7d11393bd20ee61ffc
aiecol_cameratraps.zip   MD5  af77f1a13ae56ebd8df708ea04ee7dc2
aiecol_audio.zip         MD5  373c2eca195deccd73873409e47ffe85

Generate yours with md5 -q FILE (macOS), md5sum FILE (Linux), or certutil -hashfile FILE MD5 (Windows).

Pretrained model weights

The labs download these automatically the first time you run them; we list them here so you know what's coming.

Weights	Size	Source	Used in
ResNet-18 (ImageNet)	~45 MB	torchvision auto-downloads	Labs 2, 5, 7
YOLOv8n (COCO)	~6 MB	Ultralytics auto-downloads	Labs 3, 4
YOLOv8s (COCO)	~22 MB	Ultralytics auto-downloads	Lab 4 (full run)
BirdNET v2.4	~77 MB	`birdnet` package auto-downloads	Lab 6
PNW-Cnet v5	~120 MB	`pycnet-audio` auto-downloads	Labs 7, 11

Licensing summary

Camera-trap images: CC-BY, from the LILA Oregon Critters dataset (Beery, Ruff, et al.).
Audio recordings: mixed CC-BY / CC-BY-SA / CC-BY-NC / CC0, from iNaturalist observers. Per-recording license is in audio_index.csv.
Course materials (this site, tutorials, labs): all original writing by the instructor; reuse with attribution is welcomed.
Lecture slides: see the lecture slides page — attribution to Sara Beery; please don't redistribute the files.