School of EE Seminar- speaker: Dana Cohen- A self supervised stylegan for image annotation and classification with extremely limited labels

26 בינואר 2022, 14:00 
 
School of EE Seminar- speaker: Dana Cohen- A self supervised stylegan for image annotation and classification with extremely limited labels

https://zoom.us/j/92522768275‎?pwd=MVZPSGVCKytVdmE4eTczczQwUFFiUT09

Meeting ID: 925 2276 8275

Passcode: veGm5C

 

 

Electrical Engineering Systems Seminar

                                                                                                                                             

Speaker: Dana Cohen

M.Sc. student under the supervision of Prof. Hayit Greenspan and Prof. Raja Giryes

 

Wednesday, January 26th, 2022, at 14:00

 

A self supervised stylegan for image annotation and classification with extremely limited labels

Abstract

The recent success of learning-based algorithms can be greatly attributed to the immense amount of annotated data used for training. Yet, many datasets lack annotations due to the high costs associated with labeling, resulting in degraded performances of deep learning methods. Self-supervised learning is frequently adopted to mitigate the reliance on massive labeled datasets since it exploits unlabeled data to learn relevant feature representations. In this work, we propose SS-StyleGAN, a self-supervised approach for image annotation and classification suitable for extremely small annotated datasets. This novel framework adds self-supervision to the StyleGAN architecture by integrating an encoder that learns the embedding to the StyleGAN latent space, which is well-known for its disentangled properties. The unique properties of this learned latent space enables the smart selection of representatives from the data to be labeled and used as the ground truth data for classification. This allows for improved classification performance and eliminates the necessity of using a large number of labels for adequate classification. We show that the proposed method attains strong classification results using small labeled datasets of sizes 50 and even 10. We demonstrate the superiority of our approach for the tasks of COVID-19 and liver tumor pathology identification.

 

השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז.  בצ'ט

 

 

 

 

אוניברסיטת תל אביב עושה כל מאמץ לכבד זכויות יוצרים. אם בבעלותך זכויות יוצרים בתכנים שנמצאים פה ו/או השימוש
שנעשה בתכנים אלה לדעתך מפר זכויות, נא לפנות בהקדם לכתובת שכאן >>