SensitiveImgDetect

This package provides a way to load a trained PyTorch model.

Installation

You can install this package using pip:

pip install SensitiveImgDetect

## Overview

The `Detect` class is designed for image classification tasks, allowing users to predict the class of single or multiple images using a pre-trained model. This document provides an overview of how to initialize and use the class, including example code snippets.

## Installation

Before using the `Detect` class, ensure that you have installed the necessary libraries:

```bash
pip install torch pillow torchvision

Initialization

To use the Detect class, you need to initialize it. The class constructor allows you to specify the device for computation (CPU or GPU).

from SensitiveImgDetect import Detect 

# Initialize the Detect class
detector = Detect(device='cuda')  # Use 'cpu' if you don't have CUDA

Methods

1. Detecting the Class of a Single Image

Method: detect_single_type

Description: Predicts the class label for a single image.

Parameters:

• img: An image object (PIL Image).

Returns: A string representing the predicted class label.

Example Usage:

from PIL import Image

# Load an image
img = Image.open("path_to_your_image.jpg")

# Predict the class label
predicted_label = detector.detect_single_type(img)
print(f"The predicted class for the single image is: {predicted_label}")

2. Detecting Class Probabilities for a Single Image

Method: detect_single_prob

Description: Predicts the class probabilities for a single image.

Parameters:

• img: An image object (PIL Image).

Returns: A dictionary with class labels and their corresponding probabilities.

Example Usage:

# Predict class probabilities
predicted_probs = detector.detect_single_prob(img)
print("Class probabilities for the image:")
for class_label, probability in predicted_probs.items():
    print(f"{class_label}: {probability}")

3. Detecting Classes for a List of Images

Method: detect_list_type

Description: Predicts the class labels for a list of images.

Parameters:

• img_list: A list of image objects (PIL Images).

Returns: A list of predicted class labels.

Example Usage:

# Load multiple images
images = [Image.open("image_1.jpg"), Image.open("image_2.jpg")]

# Predict class labels for the list of images
predicted_labels = detector.detect_list_type(images)
print("Predicted classes for the list of images:")
print(predicted_labels)

4. Detecting Class Probabilities for a List of Images

Method: detect_list_prob

Description: Predicts the class probabilities for a list of images.

Parameters:

• img_list: A list of image objects (PIL Images).

Returns: A list of dictionaries with class labels and their corresponding probabilities.

Example Usage:

# Predict class probabilities for the list of images
predicted_probs_list = detector.detect_list_prob(images)
for index, probs in enumerate(predicted_probs_list):
    print(f"Probabilities for image {index + 1}:")
    for class_label, probability in probs.items():
        print(f"{class_label}: {probability}")

Model Perference

Class: carton  
  Precision: 0.95  
  Recall: 0.95  
Class: other  
  Precision: 0.92  
  Recall: 0.82  
Class: politic  
  Precision: 0.79  
  Recall: 0.93  
Class: sex  
  Precision: 0.95  
  Recall: 0.88  
Overall Accuracy: 0.89  

version

model_v1

• base model

Class: carton
  Precision: 0.95
  Recall: 0.95
Class: other
  Precision: 0.92
  Recall: 0.82
Class: politic
  Precision: 0.79
  Recall: 0.93
Class: sex
  Precision: 0.95
  Recall: 0.88
Overall Accuracy: 0.89

model_v2

• add some data

Class: carton
  Precision: 0.92
  Recall: 0.88
Class: other
  Precision: 0.93
  Recall: 0.93
Class: politic
  Precision: 0.88
  Recall: 0.88
Class: sex
  Precision: 0.86
  Recall: 0.90
Overall Accuracy: 0.89