Javascript must be enabled for the correct page display

Uncertainty Disentanglement in Face Age Estimation

Ondrejka, Jakub (2024) Uncertainty Disentanglement in Face Age Estimation. Bachelor's Thesis, Artificial Intelligence.


Download (1MB) | Preview
[img] Text
Restricted to Registered users only

Download (132kB)


Face age estimation has been a popular problem in computer vision for quite some time, where the methods employed changed over time. Recognizing the distinction between apparent age and real age, apparent age estimation was chosen as a task for our model to perform. A recent development in machine learning is uncertainty disentanglement, crucial for further advancements in various tasks. However, its applications to face age estimation, particularly apparent age estimation, have yet to be explored. In order to study the performance of uncertainty disentanglement in apparent age estimation, we have employed DenseNet121 for feature extraction and implemented three methods for uncertainty estimation: Monte Carlo Dropout, Monte Carlo DropConnect, and Ensembles. Our results show that all three methods are capable of estimating apparent age, but struggle with providing appropriate uncertainties. All three methods struggle the most with the aleatoric uncertainty. It is worth mentioning that the Ensembles performed the best out of the three uncertainty disentanglement methods.

Item Type: Thesis (Bachelor's Thesis)
Supervisor name: Valdenegro Toro, M.A.
Degree programme: Artificial Intelligence
Thesis type: Bachelor's Thesis
Language: English
Date Deposited: 05 Mar 2024 13:28
Last Modified: 05 Mar 2024 13:28

Actions (login required)

View Item View Item