lyubonko/classification

Structure of the repo:

• data

  Datasets and all data-related stuff is here (could be symbolic link to the real data location)

• experiments

  Folder contains

  • [cfgs] param (config) files for experiments;
  • [logs] all intermediate results from different experiments (param files, checkpoints, logs, ...);
  • [scripts] scripts to run specific experiments;

• notebooks

  Place for jupyter notebooks. Notebooks could contain some analysis (dataset analysis, evalution results), demo, some ongoing work

• material

  Data, which include additional images, results, model weights, ...

  • [results] useful results are stored here;
  • [images] images for demo, readme, ...;
  • [weights] weights of the models (e.g. pretrained backbones )

• src

  Codebase.

Requirements

Tested with

• python 3.6
• pytorch 1.1+
• tensorboard from (tf-nightly-2.0-preview)

How to use

• load the data

(cifar10) go to the folder 'data' and run [YOU NEED ~ 300 MB]

./get_cifar10_dataset.sh

(dtd) go to the folder 'data' and run

./get_cifar10_dataset.sh

• 'train'

Run from the $PROJECT$ directory:

(cifar)

python src/main_train.py --param_file experiments/cfgs/params_cifar10.yaml

(dtd)

python src/main_train.py --param_file experiments/cfgs/params_dtd.yaml