ViT-B/16 · ImageNet
Arabica · SNI-Standard
2024 · TIP UB
Vision Transformer Research Demo

Classify Arabica Bean
Quality Instantly

Upload a photo of Arabica coffee beans and get instant quality classification based on the Indonesian National Standard (SNI) defect value system — powered by ViT-B/16 running entirely in your browser.

91.67% Accuracy No Server · 100% Client-side ViT-B/16 · ONNX 6 SNI Grades
Image Input
Click to upload or drag & drop
JPG, PNG, WEBP — any size
Preview
Try a sample — click to load
Mutu 1
mutu1
Mutu 2
mutu2
Mutu 3
mutu3
Mutu 4
mutu4
Mutu 5
mutu5
Mutu 6
mutu6
⚠ Only works with raw green Arabica beans — not roasted, ground, or brewed
Loading model from HuggingFace Hub…
Classification Result
— Awaiting classification
—
confidence
SNI Standard
Upload an image and click classify to see results.
All Class Probabilities
SNI Quality Reference
mutu1 Specialty Grade0 defects per 300g sample. Exceptional quality, no defects allowed.
mutu2 Grade 1Max 11 defect values. Excellent quality with minimal defects.
mutu3 Grade 212–25 defect values. Good quality, suitable for specialty market.
mutu4 Grade 326–44 defect values. Standard commercial quality.
mutu5 Grade 4a45–60 defect values. Below standard, limited commercial use.
mutu6 Grade 4b61–80 defect values. Low quality, significant defects present.
About This Model
Architecture
Vision Transformer B/16
ImageNet-21k pretrained
Fine-tuned on Arabica beans
Performance
91.67% prediction accuracy
Train:Test ratio 70:30
+5% vs MobileNetV2 baseline
Links
Thesis Article
GitHub Repo
HuggingFace Hub
Kaggle Dataset