❗️📢 El pròxim dia 22/12 a les 12h fem la presentació del nostre prototip ANDROMEDA!
🗓 No t’ho perdis i inscriu-te al formulari https://t.co/ALyJ4PWIwB
💡 Format: Presencial/Online
📍 Lloc: Aula Capella (ETSEIB)
#GrasptheWorld 🦾
Alumnat de l'ETSEIB i d'altres centres, apunteu-vos-hi!!
És una experiència inolbidable!
Sabeu que no us enganyo!
#femETSEIB#proETSEIB@ETSEIB_UPC@la_UPC
⚠️ OBRIM PLACES PER ENTRAR A L’EQUIP!!
Si t’interessa el món de l’enginyeria biomèdica i vols formar part d’ARM2u, et convidem a fer una entrevista!!
Contacta’ns via DM o envia’ns un mail a [email protected] i et farem arribar un formulari d’inscripció #GraspTheWorld #🦾
Practicing with machine learning, Google Collab and tensorflow! Small and understable mistakes in some characters, but very happy with the results.
All based in the article:
https://t.co/ThLOwxm44M
The algorithm:
1. Start at any cell
2. While there are non-visited cells:
- If the current cell has non-visited neighbors, choose one randomly, delete the wall in between and make it the current cell.
- Else, choose a random visited cell with one neighbor at least.
This is really neat! You take a screenshot of an equation, it gives you the LaTeX code, you can directly modify in the taskbar, copy, paste, done.
https://t.co/VMZfoNpasn
@RogerPasola Yes! Neural Style Transfer, precisely.
The basic concept behind this is a distance function that defines how different two images are. Starting from the concept image, the algorithm minimize de distance from the style image. Theoretically, the result should be in between the two
@andreuinyu El code que faig servir és una versió lleugerament modificada del del tutorial de Neural Style Transfer a la web de Pytorch, allà et pots descarregar el py
@andreuinyu Tinc el code amb Pytorch i estic fent proves en CPU però aquestes imatges han estat generades amb https://t.co/aRrJxXnPDl, on es fa servir GPU.
En CPU, en aquesta resolució, tarden ~1h
En GPU, ~5 min
Another example with an artwork of Edward Delandre (edward.delandre in Instagram)
First pic: Content image
Second pic: Style image
Last pic: Result image