Projecte llegit
Títol: Self-Supervised Learning by Image Colorization
Estudiants que han llegit aquest projecte:
- HOU, JIE (data lectura: 29-06-2021)
- Cerca aquest projecte a Bibliotècnica
Director/a: TARRÉS RUIZ, FRANCESC
Departament: TSC
Títol: Self-Supervised Learning by Image Colorization
Data inici oferta: 21-07-2020 Data finalització oferta: 21-03-2021
Estudis d'assignació del projecte:
- MU MASTEAM 2015
Tipus: Individual | |
Lloc de realització: EETAC | |
Paraules clau: | |
Deep Learning, Self-Supervised learning, Python, Image Processing, Image Recognition, Keras | |
Descripció del contingut i pla d'activitats: | |
In this thesis we will evaluate different architectures for colorization of black & white
images. Colorization is a problem that can be trained using self-supervised methodologies because it is easy to generate a huge database using completely automatic procedures, without the intervention of human annotations. We will evaluate the performance of the model and its quality in colorization. Moreover, another interesting part of the project will be to see if it is possible to use transfer learning techniques to use the same model for solving other object classification problems. Finally, the software developed should be presented in a Colab Jupyter Notebook in a tutorial form. |
|
Overview (resum en anglès): | |
Using self-supervised Learning for Grey scale image colorization.We started the experiment from YUV color space and Lab color space, optimized the models step by step based on Lab colorspace by adding an integrated fusion layer and changed the last layer to predict the color probability distribution. |