Day 4

ChatGPT-4 Image Processing

2023-10-20

Tags: GenerativeAI OpenAI ChatGPT ImageProcessing DallE3

Remember when we were all doing tutorials on Image Classification back in 2012? Using Machine Learning to answer the all important question: Is this a cat? Building training sets with thousands of images of cats and non-cats.

With the image processing features in GPT-4 those days are long gone. Now we can give GPT-4 and image and ask it to describe it. And there's no need to train it. It just works.

Prompt:

Cat / non-cat image classification Give me some images to help me with a preview image for a blog post on that topic.

Given These Images:

Describe this image:

The image shows a close-up of a black cat. The cat's mouth is slightly open, revealing its sharp teeth. The cat's nose, whiskers, and one of its eyes are also visible. The background appears to be a dimly lit room with various objects and a blue glow, possibly from a television or computer screen.

Describe this image:

A curious black cat with wide, attentive eyes peeks out from inside a gray shark-shaped pet bed. The shark bed features a mouth with white, jagged teeth, and a black eye, creating the illusion that the cat is being playfully "eaten" by the shark. Behind the shark bed, there's a plush toy resembling Hello Kitty with an orange bow, placed on top of a container adorned with a similar character design. The setting seems to be on a carpeted floor.

Describe this image:

Dynamic explosion of vibrant colored paint splatters, converging from yellow, orange, red, purple, and blue hues, against a dark background, creating a 3D-like effect with dripping and suspended droplets.

Finally, to come full circle, let's have GPT-4 describe the image of the Cat Classifier we generated at the beginning of this post.

Describe this image:

The image features a series of pastel-colored square icons connected by arrows. Each square contains a different illustration:

1. Two squares have cute white cats sitting.
2. One square has a golden cat with starry eyes.
3. One square has a grey cat-like figure on wheels.
4. There's a square with a pink heart.
5. Another square contains a brown paw print.
6. One square has a green checkmark.
7. And the last square features an abstract human silhouette or profile icon.

The overall theme of the image appears to be related to cats, with a flowchart or connectivity design element. The background color is a soft pink.

What have we learned?

With the recent advances in Generative AI and Image Classification, gone are the days of training large data sets to classify images. Now we can simply give GPT-4 an image and ask it to describe it. And it just works.