Fashion history is an integral subfield of history as a whole. Fashion plates provide important evidence of what fashion once looked like and as such are a valuable window into the lives of the people in ages past. The rise of digitization opens new avenues for aiding historians in the dating of fashion plates; following on from this, digitization also brings a greater need for artworks to be digitized and AI can be utilized in order to keep up with the demand. This provides unique challenges such as gathering data and working with a relatively limited database. Due to the lack of prior research into the subject of dating fashion plates using Artificial Intelligence (AI), said application of AI in the dating process could help future historians automate the task. Transfer learning can help streamline the dating process of fashion plates. I used several approaches with three different models (ResNet101, NasNetMobile, and InceptionV3) and achieved the best mean absolute error of 2.8 years in a range of 60 years using NasNetMobile with a simple output layer and no fine-tuning.
Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:hj-62275 |
Date | January 2023 |
Creators | Dobrzańska, Magdalena |
Publisher | Jönköping University, Jönköping AI Lab (JAIL) |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0022 seconds