LATENT SPACE MODE
The LATENT SPACE mode is a way to spatialize the frames in a two-dimensional reference frame based on their content. Here, each scene has been encoded with an artificial intelligence model and transformed into a vector composed of 512 numerical values representing the visual characteristics of the image. We propose here three methods for the spatialization of these vectors.
RANDOM
The default method. The images are simply arranged randomly in space.
PCA
Principal Component Analysis (or PCA) is a technique that simplifies complex data correlated with each other by reducing them to their most important elements. In this way, the 512 components of the initial vector are summarized in 2 numerical values associated with the X and Y axes of the reference frame.
UMAP
UMAP (Uniform Manifold Approximation and Projection) is a more complex method than PCA. While Principal Component Analysis performs linear combinations between the most expressive components of the initial vector, UMAP relies on Riemannian geometry methods to represent the topological structure and connectivity of the basic data (of dimension 512) in a lower dimension space (here 2).
Filters 
LATENT SPACE
?
0 shots found
in 0 movies in 0.00s
in 0 movies in 0.00s
MOVIE / DIRECTOR
?
Search any
director
movie
COUNTRY
?
FRAMING
?
RELEASE YEAR
?
0 films
BRIGHTNESS
?
-- %
0%
100%
MOVIE COLOR
?


both
DIALOGUE
?
TEMPERATURE
?
-- °K
1,000 °K
10,000 °K
MOVIE RATIO
?
1.37:1
Academy Ratio







