Chapter
Exciting Applications of Computer Vision and Machine Learning at Adobe
20:20 - 28:58 (08:37)
Adobe is working on exciting applications of computer vision and machine learning such as scene stitching, sky replacement, and automatic image captioning. Concept Canvas lets you assign spatial regions to keywords and uses common sense knowledge of the world to recolor the foreground of the image being edited.
Clips
Adobe is working on various applications of computer vision and machine learning such as scene stitching, sky replacement, spatial object-based image search, and automatic image captioning.
20:20 - 23:25 (03:04)
Summary
Adobe is working on various applications of computer vision and machine learning such as scene stitching, sky replacement, spatial object-based image search, and automatic image captioning. They also have projects like Project Cloak, Project Deep Fill, and Project Puppetron that focus on tasks like background removal, filling missing parts of images, and style transformation of videos and faces.
ChapterExciting Applications of Computer Vision and Machine Learning at Adobe
EpisodeGavin Miller: Adobe Research
PodcastLex Fridman Podcast
Concept Canvas allows the assignment of spatial regions to keywords in order to facilitate spatial design or layout when searching for images.
23:25 - 26:02 (02:36)
Summary
Concept Canvas allows the assignment of spatial regions to keywords in order to facilitate spatial design or layout when searching for images. This allows for more precise and creative searches and can give the user a feeling of designing rather than just searching.
ChapterExciting Applications of Computer Vision and Machine Learning at Adobe
EpisodeGavin Miller: Adobe Research
PodcastLex Fridman Podcast
In order to build common sense into AI, it would require a gradual filling-out of sets and an understanding of how an algorithm could accurately fill out an image.
26:02 - 28:58 (02:56)
Summary
In order to build common sense into AI, it would require a gradual filling-out of sets and an understanding of how an algorithm could accurately fill out an image. Additionally, algorithms should have guard rails to ensure that they only operate within their competence.