Downloads · 30 days
16
27% of all-time downloads
securepixel-convnext-v2-dual
rijal028/securepixel-convnext-v2-dual
securepixel-convnext-v2-dual is a image classification model from rijal028. Use it when you need a label for an image. The card lists the license as apache-2.0.
Sistem deteksi forensik citra resolusi dinamis tanpa kompresi paksa, dirancang untuk membedakan foto kamera asli (MS COCO / DSLR / HP) dari konten sintetis generator AI modern (Midjourney v6, SDXL, Stable Diffusion 3,…
All-time downloads
59
Public
Repo size
335 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt223 MB · 100%
At a glance
- Task
- Image Classification
- License
- apache-2.0
- Access
- Public
- Created
- Sep 19, 2026
- Updated
- Sep 20, 2026
- SHA
- 889d79e3
Model card
From the Hugging Face model README
SecurePixel: Dual-Expert ConvNeXt V2 Dynamic
Sistem deteksi forensik citra resolusi dinamis tanpa kompresi paksa, dirancang untuk membedakan foto kamera asli (MS COCO / DSLR / HP) dari konten sintetis generator AI modern (Midjourney v6, SDXL, Stable Diffusion 3, DALL-E 3).
Arsitektur
- Backbone:
convnextv2_tiny.fcmae_ft_in22k_in1k - Mekanisme Input: Resolusi Dinamis Native (tanpa downscaling ke $224 \times 224$ yang merusak derau sensor).
- Dual-Expert Layout:
guardian/: Pakar verifikasi kamera asli & resisten kompresi medsos.hunter/: Pakar penangkap diskontinuitas tekstur sintetis AI.
Hasil Validasi (5 Epoch, LR 4.0e-6)
- Total Akurasi: 86.7%
- Real Verification (Anti-False-Positive): 90.0%
- AI Catch Rate: 83.3%
- Task
- Image Classification
- License
- apache-2.0
- Languages
- en, id
- Created
- Sep 19, 2026
- Updated
- Sep 20, 2026