frankiegu / deep-virtual-try-on-with-clothes-transform

An image-based virtual try-on system with deep learning.

Deep Virtual Try-on with Clothes Transform

Source code for paper "Deep Virtual Try-on with Clothes Transform"

Overall Architecture

Step1: CAGAN

code and data

• Training: CAGAN.py

python CAGAN.py

• Testing: Testing_with_fixed_data.py

python Testing_with_fixed_data.py

• Data: MVC_image_pairs_resize_new.zip

parameters in code

Training: CAGAN.py

• Data should be put in

"./MVC_image_pairs_resize_new/1/*.jpg" (for person images)

"./MVC_image_pairs_resize_new/5/*.jpg" (for clothes images)

470: data = "data folder name"

471: train_A = "person images folder name"

473: filenames_1 = "person images folder name"

474: filenames_5 = "clothes images folder name"

617, 618: set "save model path"

Testing: Testing_with_fixed_data.py

• Data should be put in

"./MVC_image_pairs_resize_new/1/*.jpg" (for person images)

"./MVC_image_pairs_resize_new/5_test/*.jpg" (for clothes images)

215: set "model path"

220: data = "data folder name"

221: train_A = "person images folder name"

222: filenames_5 = "clothes images folder name"

224: out_root_dir = "output folder name"

225: origin_dir = "save input person images"

226: target_dir = "save target clothes images"

227: output_dir = "save output images"

228: mask_dir = "save output masks"

230: testing_number = "how much data you want to test"

Step2: Segmentation

code

https://github.com/Engineering-Course/LIP_SSL

• Modify mask: modify_mask.m

• Save the masks file to png file: show.m

• Combine all the masks: combine_with_CAGANmask.m

Step3: Transform

code and data

• Training: unet.py data.py

python unet.py

• Testing: Testing_unet.py

python Testing_unet.py

• Data: transform_data.zip transform_test_data.zip

parameters in code

Training: unet.py

336: model_dir = "save model path"

337: result_dir = "save results path"

223: set "loss type"

data.py

15: set "data path"

Testing: Testing_unet.py

16: test_data_path = "data path"

17: test_img_folder = "target clothes image folder name"

18: test_mask_folder = "mask folder name"

19: model_name = "model name"

20: result_dir = "save results path"

Step4: Combination

code

Combine_image.m

Results