Inpainting and outpainting in Prompt Engineering / GenAI - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding Inpainting Purpose

What is the main goal of inpainting in image generation?

A To increase the resolution of an image by adding pixels

B To generate a completely new image from random noise

C To fill missing or damaged parts of an image using surrounding context

D To segment an image into different object classes

💡 Hint

Think about fixing or completing parts of an existing image.

Inpainting is about filling missing or damaged areas in an image by using information from the surrounding pixels to create a seamless result.

❓ Predict Output

intermediate

Output of Outpainting Model Code

Given the following pseudocode for an outpainting model, what is the shape of the output image if the input image is 256x256 pixels and the model extends the canvas by 64 pixels on each side?

Prompt Engineering / GenAI

input_image = load_image('input.png')  # shape: (256, 256, 3)
output_image = outpaint_model(input_image, extend=64)
print(output_image.shape)

A (256, 256, 3)

B (256, 320, 3)

C (320, 320, 3)

D (384, 384, 3)

💡 Hint

Outpainting adds pixels around the original image edges.

The model extends the image by 64 pixels on each side, so width and height increase by 128 pixels total: 256 + 128 = 384.

❓ Model Choice

advanced

Best Model Architecture for Inpainting

Which model architecture is best suited for high-quality image inpainting tasks?

A Convolutional Autoencoder with skip connections

B Simple feedforward neural network

C Recurrent Neural Network (RNN)

D K-Nearest Neighbors (KNN) classifier

💡 Hint

Consider models that capture spatial features and details.

Convolutional autoencoders with skip connections preserve spatial details and context, making them ideal for inpainting.

❓ Metrics

advanced

Evaluating Outpainting Quality

Which metric is most appropriate to quantitatively evaluate the visual quality of an outpainted image compared to the original?

A Peak Signal-to-Noise Ratio (PSNR)

B Accuracy of classification labels

C Mean Squared Error (MSE) only on masked region

D Confusion matrix

💡 Hint

Think about measuring similarity between images.

PSNR measures the similarity between images and is commonly used to assess image reconstruction quality including outpainting.

🔧 Debug

expert

Debugging Inpainting Model Output

You trained an inpainting model but the output images have visible sharp edges around the filled regions, making the inpainted area obvious. What is the most likely cause?

A The input images were too large for the model

B The model was trained without using a smooth loss function or perceptual loss

C The model used too many convolutional layers

D The training dataset had too many images

💡 Hint

Consider what helps the model blend filled regions smoothly.

Using only pixel-wise loss can cause sharp edges; adding smooth or perceptual loss helps the model generate seamless transitions.