Downloads · 30 days
15
17% of all-time downloads
starlight
macroadster/starlight
starlight is a image classification model from macroadster. Use it when you need a label for an image. It is set up for transformers. The card lists the license as apache-2.0.
- Task: Detection / Extraction - Architecture: Unified CNN-based Encoder-Decoder with Residual Blocks - Input: 256x256 RGB/RGBA or metadata - Output: - Detector: sigmoid probability - Extractor: variable-length byte s…
All-time downloads
89
Public
Repo size
23.2 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx23.2 MB · 100%
At a glance
- Task
- Image Classification
- Library
- transformers
- License
- apache-2.0
- Model type
- onnx
- Access
- Public
- Created
- Nov 17, 2025
- Updated
- Nov 17, 2025
- SHA
- 66073cbe
Datasets
Model card
From the Hugging Face model README
Model Card: Starlight Unified Model 2025
Model Overview
- Task: Detection / Extraction
- Architecture: Unified CNN-based Encoder-Decoder with Residual Blocks
- Input: 256x256 RGB/RGBA or metadata
- Output:
- Detector: sigmoid probability
- Extractor: variable-length byte sequence
Training
- Dataset: Combined submissions (grok, gemini, claude, chatgpt, sample)
- Epochs: 50
- Batch Size: 16
- Optimizer: Adam
- Loss: BCE + MSE (detector), CrossEntropy (extractor)
Performance
| Metric | Value |
|---|---|
| Accuracy | 96.3% |
| AUC-ROC | 0.996 |
| F1 Score | 0.982 |
| Extraction BER | 0.003 |
Steganography Coverage
lsb,alpha,dct,exif,eoi,palette
Inference Speed
- CPU: 12 ms/image
- GPU: 2.1 ms/image
License
- Model: Apache 2.0
- Code: MIT
- Task
- Image Classification
- Library
- transformers
- Type
- onnx
- License
- apache-2.0
- Created
- Nov 17, 2025
- Updated
- Nov 17, 2025