This AI is INSANELY good at creating what you IMAGEN!! 🍓🐸
☕ BUY ME A COFFEE! ☕ https://www.buymeacoffee.com/foadmunir
Imagen can create images out of simple text input in new ways to turn it into photorealistic images that have never existed before and is even better than DALL-E2. It is a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding.
It builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation.
Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment.
----------------------------------------------------------------
follow all of these or AI will not be your friend
• Facebook - https://www.facebook.com/beyondengineering7
• Twitter - https://twitter.com/foad190
• LinkedIn - https://www.linkedin.com/in/foadmunir/
--------------------------------------------------------------------
Resources:
https://doi.org/10.48550/arXiv.2205.11487
https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html
https://developer-blogs.nvidia.com/wp-content/uploads/2022/04/Generation-with-Diffusion-Models.png
Videos:
MART PRODUCTION: https://www.pexels.com/video/a-pregnant-woman-undergoing-ultrasound-procedure-7088516/
Pat Whelen: https://www.pexels.com/video/a-person-doing-graffiti-art-by-using-a-spray-paint-5621707/
Dan Cristian Pădureț: https://www.pexels.com/video/texture-abstract-colorful-motion-4365142/
Yaroslav Shuraev: https://www.pexels.com/video/low-angle-video-of-girl-whispering-on-a-robot-8464662/
Pixabay: https://www.pexels.com/video/video-of-people-walking-855564/
OpenAI's DALL-E2 https://www.youtube.com/watch?v=qTgPSKKjfVg
Cedric Fauntleroy: https://www.pexels.com/video/woman-in-her-gray-scrub-suit-4487288/
Pavel Danilyuk: https://www.pexels.com/video/a-stressed-man-crumpling-and-throwing-paper-in-an-office-8061605/
Taryn Elliott: https://www.pexels.com/video/silhouette-of-a-woman-standing-by-the-shore-4729789/
Timecodes
0:00 - AI Insanity
0:21 Increasing Capability of AI
0:52 Some Examples of Imagen's Powers
1:05 Why It Is So Scary!
1:43 What is Imagen?
2:39 Model Overview
3:44 T5-XXL Transformer
4:09 What is an Embedding?
4:41 Diffusion
5:08 How Imagen Works and It's Architecture
6:28 Thresholding and It's Impact
6:57 How are Variations Created?
7:32 Model Evaluation
7:59 Risks and Benefits
8:51 Conclusion
Видео This AI is INSANELY good at creating what you IMAGEN!! 🍓🐸 канала Beyond Engineering
Imagen can create images out of simple text input in new ways to turn it into photorealistic images that have never existed before and is even better than DALL-E2. It is a text-to-image diffusion model with an unprecedented degree of photorealism and a deep level of language understanding.
It builds on the power of large transformer language models in understanding text and hinges on the strength of diffusion models in high-fidelity image generation.
Imagen achieves a new state-of-the-art FID score of 7.27 on the COCO dataset, without ever training on COCO, and human raters find Imagen samples to be on par with the COCO data itself in image-text alignment.
----------------------------------------------------------------
follow all of these or AI will not be your friend
• Facebook - https://www.facebook.com/beyondengineering7
• Twitter - https://twitter.com/foad190
• LinkedIn - https://www.linkedin.com/in/foadmunir/
--------------------------------------------------------------------
Resources:
https://doi.org/10.48550/arXiv.2205.11487
https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html
https://developer-blogs.nvidia.com/wp-content/uploads/2022/04/Generation-with-Diffusion-Models.png
Videos:
MART PRODUCTION: https://www.pexels.com/video/a-pregnant-woman-undergoing-ultrasound-procedure-7088516/
Pat Whelen: https://www.pexels.com/video/a-person-doing-graffiti-art-by-using-a-spray-paint-5621707/
Dan Cristian Pădureț: https://www.pexels.com/video/texture-abstract-colorful-motion-4365142/
Yaroslav Shuraev: https://www.pexels.com/video/low-angle-video-of-girl-whispering-on-a-robot-8464662/
Pixabay: https://www.pexels.com/video/video-of-people-walking-855564/
OpenAI's DALL-E2 https://www.youtube.com/watch?v=qTgPSKKjfVg
Cedric Fauntleroy: https://www.pexels.com/video/woman-in-her-gray-scrub-suit-4487288/
Pavel Danilyuk: https://www.pexels.com/video/a-stressed-man-crumpling-and-throwing-paper-in-an-office-8061605/
Taryn Elliott: https://www.pexels.com/video/silhouette-of-a-woman-standing-by-the-shore-4729789/
Timecodes
0:00 - AI Insanity
0:21 Increasing Capability of AI
0:52 Some Examples of Imagen's Powers
1:05 Why It Is So Scary!
1:43 What is Imagen?
2:39 Model Overview
3:44 T5-XXL Transformer
4:09 What is an Embedding?
4:41 Diffusion
5:08 How Imagen Works and It's Architecture
6:28 Thresholding and It's Impact
6:57 How are Variations Created?
7:32 Model Evaluation
7:59 Risks and Benefits
8:51 Conclusion
Видео This AI is INSANELY good at creating what you IMAGEN!! 🍓🐸 канала Beyond Engineering
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Fast Image Editing with Muse AI!CREATE monsters and cities with IMAGEN!Ingenuity...FIVE Things You Must Know About The Mars Helicopter #ShortsElevate Your Image Editing Skills With Muse AI!Falcon Legs Let This Drone LAND Anywhere!Large Language Models Are Coming After Google Search #shortsHow Football’s Offside AI worksMuse vs DALL-E2 vs ImagenEngineering Failures Behind Titan’s Implosion!This UNIQUE Orbit is our GATEWAY to the MOON!This AI Is Flawless!Inside Las Vegas Sphere…Immersive Audio On ANOTHER Level!What Happens When The Electric Grid Goes DOWN!How BOEING Lost It's Way...How To Catch A Rocket With A Helicopter!ChatGPT Is Changing The WAY We Work Forever!Weird Things That Break Muse AIHow Good Is Tesla Semi?Learn Maxwell Equations with ChatGPT!Did Earth's Inner Core STOP Spinning!?