Nepal_Vehicle_License_Plates_Detection_Version3

<repository> <title>📌 Nepal Vehicle License Plates Detection Model (Version 3)</title> <section> <heading>🔍 Description</heading> <content> **Version 3** of the Nepal Vehicle License Plates Detection Model offers advanced functionality to detect **individual characters** in Nepalese vehicle license plates and draw bounding boxes around each one. This version enhances the accuracy and granularity of detection compared to earlier versions, making it ideal for real-world applications. </content> </section> <section> <heading>✨ Key Features</heading> <features> <feature> <name>📝 **Character-Wise Detection**</name> <summary>Detects and draws bounding boxes for each character in a license plate for precise recognition.</summary> </feature> <feature> <name>🎯 **High Precision and Recall**</name> <summary>Achieves industry-leading metrics, ensuring accurate and reliable performance.</summary> </feature> <feature> <name>⚡ **Real-Time Inference**</name> <summary>Optimized for fast predictions, making it suitable for live detection applications.</summary> </feature> </features> </section> <section> <heading>📜 Earlier Model</heading> <content> Explore the earlier version of this project, which detects license plates as a whole: [Nepal Vehicle License Plate Detection](https://huggingface.co/krishnamishra8848/Nepal-Vehicle-License-Plate-Detection) </content> </section> <section> <heading>📊 Model Performance Metrics</heading> <metrics> <metric> <name>**Precision**</name> <value>0.985</value> <description>Percentage of correct bounding box predictions among all predictions.</description> </metric> <metric> <name>**Recall**</name> <value>0.984</value> <description>Percentage of ground truth objects successfully detected.</description> </metric> <metric> <name>**mAP@50**</name> <value>0.994</value> <description>Mean Average Precision at IoU threshold 0.5.</description> </metric> <metric> <name>**mAP@50-95**</name> <value>0.861</value> <description>Mean Average Precision across multiple IoU thresholds (0.5 to 0.95).</description> </metric> <metric> <name>**Inference Speed**</name> <value>~2.4ms per image</value> <description>Time taken to process a single image during inference.</description> </metric> </metrics> </section> </repository>

# Example Code: Test this code on colab
# Step 1: Install required libraries
!pip install ultralytics
!pip install huggingface_hub

# Step 2: Download the model from Hugging Face
from huggingface_hub import hf_hub_download

# Replace "krishnamishra8848/Nepal_Vehicle_License_Plates_Detection_Version3" with your model's Hugging Face repo ID
repo_id = "krishnamishra8848/Nepal_Vehicle_License_Plates_Detection_Version3"
model_path = hf_hub_download(repo_id=repo_id, filename="best.pt")

print(f"Model downloaded and saved at: {model_path}")

# Step 3: Load the model
from ultralytics import YOLO
model = YOLO(model_path)  # Load the model

# Step 4: Upload an image for inference
from google.colab import files
import matplotlib.pyplot as plt
import cv2

# Allow user to upload an image
uploaded = files.upload()

# Get the uploaded file's path
image_path = list(uploaded.keys())[0]

# Step 5: Perform inference
results = model.predict(source=image_path)

# Step 6: Display the results
# Extract the annotated image
annotated_image = results[0].plot()

# Display the image with bounding boxes
plt.figure(figsize=(10, 10))
plt.imshow(cv2.cvtColor(annotated_image, cv2.COLOR_BGR2RGB))
plt.axis("off")
plt.show()