In the wake of DALL-E, the artificial intelligence image generator designed by OpenAI, many tools have emerged that use AI to produce images from textual queries. The results have often proven to be stunning, raising questions about our ability to distinguish truth from falsehood. These concerns are not new, they were already very present when deepfakes appeared a few years ago. As time goes by, AIs will undoubtedly develop and become more difficult to unmask. In this article, we offer you some tips to spot images produced by AI!
Here are some tools to combine to distinguish artificially produced images.
1- Check if the use of an AI is indicated
While it may be tempting to hide the fact that an image was generated by artificial intelligence, some users are honest enough to indicate the actual source of their image. In some cases, the objective of the operation is to demonstrate the performance of the tool, in which case it is necessary to indicate on which software the image was produced. The first thing to do is to check if a credit is present on the image. If it is posted on a social network, this credit can be indicated in the comments of the publication. Some image generators even propose a symbol to sign their work. DALL-E, for example, adds a small multicolored watermark at the bottom right of the images (see photo above). On its side, Craiyon integrates its logo in the same place.
2- Look for inconsistencies or distortions in the image
Artificial intelligence are far from perfect, and most still make gross errors. Some elements of the image can be blurred, scrambled or disproportionate, and it is not uncommon to see faces without noses or hands with 6 fingers. Human representations seem to be the most difficult to reproduce: teeth and (especially) eyes are often distorted in deepfake images or videos.
If you want to train yourself to spot faces created by AI, you can play the game which face is real, in which you have to guess which of the two faces really exists. You’ll see, it’s not very difficult!
3- Check if the elements of the image are consistent with each other
In many cases, artificial intelligences produce images that look quite good, but in which the elements do not work well together. For example, the different components of an image may have slightly different drawing styles.
So here are some details we suggest you analyze :
- Are the shadows and lighting realistic?
- Are the colors in the image consistent with each other?
- Do the different elements have the same style or level of realism?
4- Use image recognition software
As for textual analysis (following the emergence of ChatGPT), some solutions are being implemented to detect whether an image has been produced by artificial intelligence. At this time, not all of them are efficient: the GAN detector from Mayachitra Inc, a company initially specialized in image recognition by AI, still has some way to go to be reliable. But these tools will gain to be perfected by being trained. In addition, many renowned companies, such as Microsoft or Intel, have been working for a few years on deepfakes recognition, and their expertise could prove useful in the development of new tools. For its part, the company Hive AI offers companies an API to detect and moderate AI-generated artworks.
5- Distinguish between AI styles
At first glance, artificial intelligence image generators are supposed to be able to reproduce a wide variety of styles. It is indeed possible to ask them to create an image in a particular style of illustration, or by reproducing the touch of a great painter. However, it can be seen that, by default, the tools tend to adopt a particular style that can be recognized by the trained eye. By producing images yourself on the various generators, you will train yourself to identify drawing registers or color palettes that recur regularly.
Top 5 AI Image Generator Tools
- StyleGAN : StyleGAN is a state-of-the-art AI image generator that is capable of producing high-resolution images that look like they were taken by a professional photographer. It can generate human faces, animals, landscapes, and many other types of images.
- DALL-E : DALL-E is an AI image generator created by OpenAI that can generate images from textual descriptions. It can create images of anything that can be described in words, including animals, objects, and scenes.
- GPT-3 Image Generation : GPT-3 is a natural language processing tool that can also generate images based on text prompts. It is not as sophisticated as some of the other image generators on this list, but it is still capable of producing impressive results.
- BigGAN : BigGAN is an AI image generator that was created by researchers at Google. It can generate high-quality images of a wide range of objects, animals, and scenes.
- CycleGAN : CycleGAN is an AI image generator that is capable of translating images from one style to another. For example, it can take a photograph of a winter landscape and transform it into a summer landscape. It is useful for tasks like style transfer and image editing.
RELATED POSTS :