Summer Project by junzhi Ning

This is a organized github repository for the Scientific Research Project in SCIE30001 at UOM

项目分析半监督与无监督迁移学习计算机视觉算法异同点，提出预训练神经网络对模型影响的可能性

Dependencies:

• Language: i.e Python 3.7.11
• Packages / Libraries:
  • pandas
  • torch>=1.7.0
  • torchvision
  • qpsolvers
  • numpy
  • prettytable
  • tqdm
  • scikit-learn
  • webcolors
  • matplotlib
  • torch>=1.7.0
  • Transfer library github from https://github.com/thuml/Transfer-Learning-Library
  • libraries from https://github.com/YBZh/Bridging_UDA_SSL and https://github.com/YBZh/Bridging_UDA_SSL

Datasets

• External dataset 1: Face-blurred ILSVRC2012–2017 classification -> https://image-net.org/download-images.php
• External dataset 2: Office31 Dataset -> https://faculty.cc.gatech.edu/~judy/domainadapt/#datasets_code

PS: Some datasets are not used in the report.

Directory

• data: Contain All preprocessed files and small external datasets supporting the analysis.Usually Store on the local machine to avoid occupying the online storage
• plots_results: All plots and outputresults both for model training and reporting writting. This file also contains the storing excel files for data outputing used in the report.
• code: Contain all code implementations in this project
  • testing3: Designed for initial debugging purposes
  • version1.0: Designed for Alexnet model pretraining with selected classes and labels masked
  • version2.0: Designed for UDA fine tuning on pretrained models
  • version3.0: Designed for SSL fine tuning on pretrained models
• reports_and_presentation : Storing the old versions of report and presentation slides for archived purposes.

References:

See the references in the report link