AIML: Exercise 2.B: Popular datasets - ImageNet

ImageNet Dataset

The ImageNet project is a large visual database for visual object recognition software research. The idea for this project was conceived over 15 years ago by AI researcher Fei-Fei Li. The ImageNet team presented their dataset for the first time in 2009.

Keras comes bundled with many pre-trained classification models. As of Keras version 2.11, there are 19 different pre-trained models available, where some versions contain many variants as well. The list of models can be found here. Here we will use the following pre-trained models to make predictions on several sample test images.

- VGG16

- ResNet50

- InceptionV3

AI Image Recognition is the process of using artificial intelligence to identify and categorize objects within an image, a task that, while intuitive for humans, is complex for machines due to the significant processing power required.

Here's a simple Jupyter Notebook exercise for students to perform image classification using a pre-trained model on the ImageNet dataset. This exercise will guide them through loading a pre-trained model, making predictions, and visualizing the results.

Exercise: Image Classification with ImageNet

Step 1: Setup

First, ensure you have the necessary libraries installed. You can install them using pip if you haven't already:

%pip install torch torchvision matplotlib

Step 2: Import Libraries

import torch
import torchvision.transforms as transforms
from torchvision import models
from PIL import Image
import matplotlib.pyplot as plt
import json

Step 3: Load Pre-trained Model

Load a pre-trained model (e.g., ResNet-18) and set it to evaluation mode:

import ssl
ssl._create_default_https_context = ssl._create_unverified_context

model = models.resnet18(pretrained=True)
model.eval()

Step 4: Load and Preprocess Image

Load an image and apply the necessary transformations:

def preprocess_image(image_path):
    input_image = Image.open(image_path)
    preprocess = transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor(),
        transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),
    ])
    input_tensor = preprocess(input_image)
    input_batch = input_tensor.unsqueeze(0)
    return input_batch

import urllib.request

# Download a sample image
image_url = 'https://erasmus.tsp.edu.rs/wp-content/uploads/2024/12/rcd1024.jpg'  # Replace with a valid image URL
image_path = 'rcd1024.jpg'
urllib.request.urlretrieve(image_url, image_path)

input_batch = preprocess_image(image_path)

Step 5: Make Predictions

Pass the preprocessed image through the model to get predictions:

with torch.no_grad():
    output = model(input_batch)

Step 6: Decode Predictions

Download the ImageNet class labels and decode the predictions:

import urllib.request

# Download the labels file
labels_url = 'https://raw.githubusercontent.com/anishathalye/imagenet-simple-labels/master/imagenet-simple-labels.json'
labels_path = 'imagenet-simple-labels.json'
urllib.request.urlretrieve(labels_url, labels_path)

# Load the labels
with open(labels_path) as f:
    labels = json.load(f)

# Get the predicted label
_, predicted_idx = torch.max(output, 1)
predicted_label = labels[predicted_idx.item()]

print(f'Predicted label: {predicted_label}')

Predicted label: power drill

Step 7: Visualize the Image and Prediction

Display the image along with the predicted label:

def show_image(image_path, label):
    image = Image.open(image_path)
    plt.imshow(image)
    plt.title(f'Predicted: {label}')
    plt.axis('off')
    plt.show()

show_image(image_path, predicted_label)

Instructions for Students

1. Follow the steps in the notebook to load and preprocess an image.

2. Use the pre-trained ResNet-18 model to make predictions.

3. Decode the predictions and display the image with the predicted label.

4. Experiment with different images and observe the model's performance.

Completion requirements:

Make a submission