# DeepFashionNetwork

**Repository Path**: jane-one/DeepFashionNetwork

## Basic Information

- **Project Name**: DeepFashionNetwork
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2020-10-30
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

 Data---
 This model is trained using the DeepFashion database
 https://drive.google.com/drive/folders/0B7EVK8r0v71pWGplNFhjc01NbzQ
 In case the above link is removed, try navigate through the follwing link
 http://mmlab.ie.cuhk.edu.hk/projects/DeepFashion.html
 
 Models---
 model_w_label: this model only trains the cloth category directly
 model_w_bbox_n_label: this model aims to train a bounding box area first and use that to assist the category training. 
  
 Instructions---
 For detailed instructions on data preparation of LMDB and HDF5 check under Docs/ 
 Once you got the data prepared, here's what you need for training. 
 
 Training---
 Eg.
 train with ResNet-50 weight initialization(optional) and subtrain1 data set, and save the log
 ../../../../build/tools/caffe train -solver solver.prototxt -weights ResNet-50-model.caffemodel  2>&1 | tee subtrain1.log 
 Note: The ResNet-50 weight file is too big to upload. Refer to He et al.'s ResNet git repo listed under 'Reference'

Testing---
Eg.
test with subtrain2 data set and save log file (subtrain2 has 52292 data points and 6536.5 batch, test itself round to 6535)
../../../../../build/tools/caffe test -model train.prototxt -weights caffenet_train_iter_60000.caffemodel -gpu 0 -iterations 6537 2>&1 | tee subtrain2.log