中文文档 (Chinese) | Release Notes
Smart AI Culling Tool for Bird Photographers
Shoot freely, cull easily! A smart photo culling software designed specifically for bird photographers. It uses multi-model AI technology to automatically identify, rate, and filter bird photos, significantly improving post-processing efficiency.
To run SuperPicky from source or build it yourself, you must first download the required AI models:
git clone https://github.com/jamesphotography/SuperPicky.git
cd SuperPicky
pip install -r requirements.txt
python scripts/download_models.py- YOLO11 Detection: Precise bird detection and segmentation masks.
- SuperEyes: Detects eye visibility and calculates head sharpness.
- SuperFlier: Identifies flight poses for bonus points.
- TOPIQ Aesthetics: Assesses overall image aesthetics, composition, and lighting.
| Stars | Condition | Meaning |
|---|---|---|
| ⭐⭐⭐ | Sharpness OK + Aesthetics OK | Excellent, worth editing |
| ⭐⭐ | Sharpness OK OR Aesthetics OK | Good, consider keeping |
| ⭐ | Bird found but below threshold | Average, usually delete |
| 0 | No bird / Poor quality | Delete |
Automatically set thresholds based on your experience:
- 🐣 Beginner: Sharpness>300, Aesthetics>4.5 (Keep more)
- 📷 Intermediate: Sharpness>380, Aesthetics>4.8 (Balanced)
- 👑 Master: Sharpness>520, Aesthetics>5.5 (Strict)
- Pick (Flag): Top 25% intersection of sharpness & aesthetics among 3-star photos.
- Flying: Green label for bird-in-flight photos.
- Exposure: Filters over/under-exposed shots (Optional).
- Sort by Stars: Auto-move to 0star/1star/2star/3star folders.
- EXIF Write: Writes ratings, flags, and scores to RAW metadata.
- Lightroom Compatible: Sort and filter immediately after import.
- Undo: One-click reset to restore original state.
Apple Silicon (M1/M2/M3/M4) (v4.2.0)
- GitHub Download
- Google Drive (Mirror)
- Baidu Netdisk Code: t6c4
- Quark
Intel (Pre-2020 Mac) (v4.2.0)
- GitHub Download
- Google Drive (Mirror)
- Baidu Netdisk Code: 3821
- Quark
CUDA-GPU Version (v4.2.0 Beta)
- Baidu Netdisk Code: c2a6
- Google Drive (Mirror)
- Quark
CPU Version (v4.2.0)
- GitHub Download
- Baidu Netdisk Code: 872v
- Google Drive (Mirror)
- Quark
- Select Folder: Drag & drop or browse for a folder with bird photos.
- Adjust Thresholds (Optional): Sharpness (200-600), Aesthetics (4.0-7.0).
- Toggle Features: Flight detection, Exposure check.
- Start: Click to begin AI processing.
- Review: Photos are organized; import to Lightroom to see ratings.
- 🚀 Architecture: SQLite migration, ~1.9x speedup.
- 🌟 Community: Thanks @OscarKing888 for Sony ARW & UTF-8 fixes.
- 🧹 Clean: Unified temp files to hidden cache dir.
- 🔧 Fixes: Chinese path support, ExifTool deadlock, Plugin metadata.
SuperPicky supports multiple English naming standards for bird species via the AviList v2025 mapping. Choose your preferred format in Settings > Culling Criteria > Species Name Format:
| Format | Source |
|---|---|
| Default (OSEA Model) | Original model training names |
| AviList v2025 | AviList unified English names |
| Clements / eBird v2024 | Cornell/eBird taxonomy |
| BirdLife v9 | BirdLife International |
| Scientific Name Only | Binomial nomenclature |
Updating AviList: The mapping is built from AviList-v2025-11Jun-extended.xlsx (located in scripts_dev/) using an offline build script. When a new AviList version is released (typically annually), replace the xlsx file in scripts_dev/ and re-run:
pip install openpyxl # first time only
python scripts_dev/build_avilist_mapping.pyReview unmatched species in the report output and add manual overrides to scripts_dev/avilist_manual_overrides.json if needed.
Open sourced under GPL-3.0 License.
This project uses:
- YOLO11 by Ultralytics
- OSEA by Sun Jiao (github.com/sun-jiao/osea)
- SuperEyes (SuperBirdEye) by Jordan Yu (于若君)
- SuperFlier by Jordan Yu (于若君)
- TOPIQ by Chaofeng Chen et al.
- AviList: AviList Core Team. 2025. AviList: The Global Avian Checklist, v2025. https://doi.org/10.2173/avilist.v2025 — Licensed under CC BY 4.0