Th-Slip-OCR-K

💳 Th-Slip-OCR-K: Precision Financial Detection

A lightweight, high-velocity object detection model engineered to extract full telemetry from Kasikorn Bank (K-Bank) transaction slips.

---

⚡ Overview

Th-Slip-OCR-K is a specialized computer vision model designed as a pre-processing engine for OCR pipelines. It isolates specific regions of interest within financial documents, eliminating noise and enabling higher accuracy for text extraction systems.

🚀 V2 Update: Enhanced detection capabilities to cover sender/receiver names and masked account numbers.

• Architecture: Ultralytics YOLOv8 Nano (v8.3.0)
• Training Hardware: Apple M4 Silicon (Metal Performance Shaders)
• Optimization: Low-latency inference, suitable for edge devices and mobile deployment.

🎯 Detection Classes (Updated)

The model is now trained to identify 6 critical zones with high confidence:

💻 Usage (Python)

Seamlessly integrate into your workflow using the ultralytics library.

from ultralytics import YOLO

# 1. Load the model directly from Hugging Face Hub
model = YOLO("https://huggingface.co/Pondet/Th-Slip-OCR-K/resolve/main/best.pt")

# 2. Run Inference
results = model("path/to/your/slip.jpg")

# 3. Display Results
results[0].show()

📊 Performance Metrics

Evaluated on the latest K-Bank validation set (103 samples):

• mAP50: 0.995 (99.5%)
• Precision: 0.998 (High fidelity)
• Recall: 1.0 (Zero miss rate)

⚠️ Scope & Limitations

• Domain Specific: Optimized exclusively for Kasikorn Bank (Green Theme) slip layouts.
• Privacy Aware: The model detects masked account numbers (x) accurately but cannot reveal hidden digits without external database verification.
• Educational Purpose: Developed as a Proof-of-Concept for automated household accounting.

---

Maintained by Pondet