Image understanding and description in Prompt Engineering / GenAI - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load an image using PIL.

Prompt Engineering / GenAI

from PIL import Image
img = Image.[1]('example.jpg')

Drag options to blanks, or click blank then click option'

Aopen

Bimport

Cread

Dload

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'load' or 'read' instead of 'open' causes errors.

Trying to use 'import' as a method.

The Image.open function loads an image file into memory for processing.

2fill in blank

medium

Complete the code to convert an image to a tensor for model input.

Prompt Engineering / GenAI

import torchvision.transforms as transforms
transform = transforms.Compose([
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
tensor_img = transform([1])

Drag options to blanks, or click blank then click option'

Ainput_data

Bimage_path

Cimage_tensor

Dimg

Attempts:

3 left

💡 Hint

Common Mistakes

Passing a file path string instead of an image object.

Passing an already tensor variable.

The variable img holds the loaded image, which we convert to a tensor.

3fill in blank

hard

Fix the error in the code to generate image captions using a pretrained model.

Prompt Engineering / GenAI

from transformers import VisionEncoderDecoderModel, ViTImageProcessor, AutoTokenizer
model = VisionEncoderDecoderModel.from_pretrained('nlpconnect/vit-gpt2-image-captioning')
processor = ViTImageProcessor.from_pretrained('nlpconnect/vit-gpt2-image-captioning')
tokenizer = AutoTokenizer.from_pretrained('nlpconnect/vit-gpt2-image-captioning')

pixel_values = processor(images=img, return_tensors='pt').[1]()
output_ids = model.generate(pixel_values)
caption = tokenizer.decode(output_ids[0], skip_special_tokens=True)
print(caption)

Drag options to blanks, or click blank then click option'

Asqueeze

Bunsqueeze

Cflatten

Dreshape

Attempts:

3 left

💡 Hint

Common Mistakes

Using squeeze removes dimensions and causes errors.

Using flatten or reshape changes tensor shape incorrectly.

The processor output needs a batch dimension, so unsqueeze() adds it.

4fill in blank

hard

Fill both blanks to create a dictionary of image features and their lengths.

Prompt Engineering / GenAI

features = {img_id: [1] for img_id, img in images.items() if len([2]) > 0}

Drag options to blanks, or click blank then click option'

Aprocessor(images=img, return_tensors='pt').pixel_values

Bimg

Cimg_id

Dimages

Attempts:

3 left

💡 Hint

Common Mistakes

Using img_id or images instead of the image object.

Not using the processor to get pixel values.

We extract pixel values from each image and check the length of the image object.

5fill in blank

hard

Fill all three blanks to filter captions longer than 5 words and create a summary dictionary.

Prompt Engineering / GenAI

summary = {img_id: caption for img_id, caption in captions.items() if len(caption.[1]()) > [2] and caption.[3](' ') > 0}

Drag options to blanks, or click blank then click option'

Asplit

B5

Ccount

Dstrip

Attempts:

3 left

💡 Hint

Common Mistakes

Using strip instead of split for word count.

Comparing length to wrong number.

Using count with wrong character.

We split captions into words, check if length is greater than 5, and count spaces to ensure words exist.