Muhammad Firdho
commited on
Commit
•
53aa608
1
Parent(s):
b527bc5
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# 🎯 # Image Classification Model for Medical Waste Classification
|
2 |
+
|
3 |
+
This is an image classification model trained to classify medical waste into 4 categories, namely cytotoxic, infectious, pathological, and pharmaceutical. The model is based on the Inception v3 architecture and has been adapted to a specific dataset for the task of medical waste classification.
|
4 |
+
|
5 |
+
# 🎯 Model Description
|
6 |
+
|
7 |
+
The model is based on the Inception v3 architecture with modifications to the fully connected layers for adapting it to the specific image classification task. The architecture consists of a feature extractor followed by a global average pooling layer and fully connected layers with ReLU activation and dropout.
|
8 |
+
|
9 |
+
# 🎯 Usage
|
10 |
+
|
11 |
+
You can use the model that I have saved in pt format as follows:
|
12 |
+
|
13 |
+
```python
|
14 |
+
|
15 |
+
import torch
|
16 |
+
from torchvision import transforms
|
17 |
+
from PIL import Image
|
18 |
+
import matplotlib.pyplot as plt
|
19 |
+
import numpy as np
|
20 |
+
|
21 |
+
def predict_image(image_path, model, transform, class_names):
|
22 |
+
# Load the image
|
23 |
+
image = Image.open(image_path)
|
24 |
+
# Apply transformations
|
25 |
+
image = transform(image).unsqueeze(0) # Add batch dimension
|
26 |
+
|
27 |
+
# Set the model to evaluation mode
|
28 |
+
model.eval()
|
29 |
+
|
30 |
+
# Make predictions
|
31 |
+
with torch.no_grad():
|
32 |
+
outputs = model(image.to(device))
|
33 |
+
_, predicted = torch.max(outputs, 1)
|
34 |
+
predicted_class = predicted.item()
|
35 |
+
predicted_label = class_names[predicted_class]
|
36 |
+
probabilities = torch.softmax(outputs, dim=1)[0]
|
37 |
+
confidence = probabilities[predicted_class].item()
|
38 |
+
return predicted_class, predicted_label, confidence
|
39 |
+
|
40 |
+
# Define transformation to be applied to the input image
|
41 |
+
image_transform = transforms.Compose([
|
42 |
+
transforms.Resize((299, 299)), # Resize to match InceptionV3 input size
|
43 |
+
transforms.ToTensor(),
|
44 |
+
# You can add more transformations such as normalization if needed
|
45 |
+
])
|
46 |
+
|
47 |
+
# Load the trained model
|
48 |
+
model = torch.load('__directory where you save the model__')
|
49 |
+
model.to(device)
|
50 |
+
|
51 |
+
# Load class names (assuming you have a list of class names)
|
52 |
+
class_names = ['cytotoxic', 'infectious', 'pathological', 'pharmaceutical']
|
53 |
+
|
54 |
+
# Provide the path to the image you want to predict
|
55 |
+
image_path = '__the directory where you store the images you want to classify__'
|
56 |
+
|
57 |
+
# Load the true label (assuming you have it)
|
58 |
+
true_label = 'pathological'
|
59 |
+
|
60 |
+
# Predict the class label
|
61 |
+
predicted_class, predicted_label, confidence = predict_image(image_path, model, image_transform, class_names)
|
62 |
+
|
63 |
+
# Display the image
|
64 |
+
image = Image.open(image_path)
|
65 |
+
plt.imshow(np.array(image))
|
66 |
+
plt.axis('off')
|
67 |
+
plt.title(f'True Class: {true_label} \n Predicted Class: {predicted_label} (Confidence: {confidence*100:.2f}%)')
|
68 |
+
plt.show()
|
69 |
+
|
70 |
+
```
|