2023-11-08 Meeting #2

 

SLR: https://pushy-rodent-1d9.notion.site/SLR-Tabula-6b17b591e6d040cf8ef2bbda92098207

 

Izveidot jaunas 2 SLR tabulas:

  1. Aging effect on face re-identification

    1. https://www.researchgate.net/publication/352113886_Deep_Face_Age_Progression_A_Survey

      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8397612/

      https://paperswithcode.com/dataset/morph

      https://sci-hub.hkvisa.net/10.1016/j.patcog.2017.10.015#google_vignette

      https://dcnhan.github.io/projects/aging_project/the-agfw-database.html

      https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8597988/

      https://xiatian-zhu.github.io/papers/ChengEtAl_PR2020.pdf

      https://www.sciencedirect.com/science/article/abs/pii/S0031320317304041

       

  2. Generative models for face aging (datu kopu augmentācijas metode)

    1. https://www.perplexity.ai/search/447decbc-b91c-47fe-a43b-c4a7c0a18b5b?s=u

Pareizā FaceNet implementācija: https://github.com/tamerthamoqa/facenet-pytorch-vggface2

Nepareizā FaceNet implementācija https://github.com/timesler/facenet-pytorch

Metodes iedalās

  1. Klasifikācija - Jāzin visas atpazīstamās sejas apmācību laikā

  2. Embedding - Nav Jāzin visas atpazīstamās sejas apmācību laikā - FaceNet, DML Triplet LossEfekts uz vecumu

 

Embedding piemērs:

image-20231108191420569