Skip to Main Content

AI Images: Creation and Analysis

Image description

These tools allow you to upload an image and will offer a description. Depending on the image they will offer context or analysis. Very useful for writing alt-text. 

Example

Meme using the 2021 Suez Canal incident photo. Shows the massive Ever Given container ship stuck in the canal, labeled 'The conference paper abstract', while a tiny bulldozer at its base is labeled 'The actual conference paper'. The contrast between the enormous ship and small bulldozer illustrates the common disconnect between ambitious academic paper abstracts and their more modest content. The photo shows the ship's green and red hull against a sandy desert background.This is a humorous meme that uses the famous Ever Given container ship incident in the Suez Canal as a metaphor for academic writing. The image shows the massive container ship (labeled "The conference paper abstract") stuck in the canal, while a tiny bulldozer (labeled "The actual conference paper") attempts to dislodge it at the base.

The joke plays on the common academic experience where paper abstracts can be very grand and ambitious in their scope and claims, while the actual paper content may be more modest in comparison. The stark size difference between the enormous ship and the small bulldozer perfectly illustrates this humorous disconnect.

The photo itself is from the March 2021 Suez Canal obstruction incident, where the Ever Given container ship became stuck and blocked global shipping for several days. The background shows the sandy banks of the canal, and you can see the ship's distinctive green hull and red bottom.

(Claude.ai)

Object detection

Object detection identifies and locates objects within an image by drawing bounding boxes around them. It is widely used for security surveillance, self-driving cars, and retail analytics. Object detection tools are mostly designed for developers and those with programming knowledge. 

Black-and-white security camera footage of a garage with various objects inside. Two children's toy cars are outlined with orange bounding boxes, labeled as 'car: 82%' and 'car: 76%' by an object detection system. A larger plastic ride-on toy, a sled, a long-handled tool, and a small storage rack are also visible. The garage door in the background is slightly open. The timestamp in the top right corner reads 'Dec 10, 2024, 03:50:44 PM.

A still from a security camera in a garage, where the system has identified two "cars".

You can try object detection using this image sonification tool created by Both Rocks!. The tool's purpose is to create a sound file from an image but it does object detection as part of its process. You can find instructions below the link. 

Image segmentation

Image segmentation is a process in computer vision that involves dividing an image into multiple segments or regions. The goal is to simplify or change the representation of an image into something more meaningful and easier to analyze. Here are the main types of image segmentation:

  1. Semantic Segmentation: This type assigns a label to each pixel in the image, classifying it into a predefined category. For example, in an image of a street scene, all pixels belonging to cars might be labeled as "car," and all pixels belonging to the road might be labeled as "road."

  2. Instance Segmentation: This goes a step further than semantic segmentation by not only classifying each pixel but also distinguishing between different instances of the same object. For example, it can differentiate between two separate cars in the same image.

  3. Panoptic Segmentation: This combines both semantic and instance segmentation, providing a complete understanding of the scene by labeling each pixel with both a class and an instance identifier.

Image segmentation is widely used in various applications, such as medical imaging (e.g., tumor detection), autonomous driving (e.g., identifying road signs and obstacles), and image editing (e.g., background removal).

Image inpainting

Inpainting is a technique used in image processing to restore missing or damaged parts of an image. The goal is to fill in these areas in a way that is visually consistent with the surrounding content. Here are some common applications of inpainting:

  1. Restoration of Old Photos: Repairing scratches, tears, and other damage in historical photographs.
  2. Object Removal: Removing unwanted objects or people from an image and filling in the background seamlessly.
  3. Image Editing: Enhancing images by filling in gaps or extending backgrounds.

Inpainting algorithms use various methods to predict and generate the missing parts, such as texture synthesis, patch-based methods, and deep learning techniques.

Example of a damaged fresco modified with Stable Diffusion Inpainting.

Image similarity

Image similarity in computer vision refers to the process of determining how similar two images are to each other. This can be useful in various applications, such as image retrieval, duplicate detection, and clustering. Image similarity is widely used in applications like content-based image retrieval, where you want to find images similar to a query image, and in detecting duplicates or near-duplicates in large image datasets.It can also be used to analyze an image dataset and create cluster of similar images. 

McGill LibrariesQuestions? Ask us!
Privacy notice