search

layumi / AICIty-reID-2020

:red_car: The 1st Place Submission to AICity Challenge 2020 re-id track (Baidu-UTS submission)
MIT License
455 stars 108 forks source link

USE of UNIT (cycle GAN) #26

Open anmol19005 opened 4 years ago

anmol19005 commented 4 years ago

I was going through your research paper and was parallelly going through the codes. I was not able to find how you used UNIT(cycle gan) to perform style transfer and then content manipulation. Kindly help us understand the flow of your code or provide a link where you have trained synthetic images on cycleGAN.

layumi commented 4 years ago

Hi @anmol19005

Thanks a lot for your attention on our work. Sorry. We do not provide the cyclegan code. Actually, you can prepare data and train it easily via the official code (https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).

If you are interested in UNIT-like framework with generation and re-id, you may refer to my previous work DG-Net https://github.com/NVlabs/DG-Net

anmol19005 commented 4 years ago

Hi Zhedong,

I wanted to know about the distribution of validation (query) and gallery dataset (like class distribution), used for reranking. Also have you used Image cloning? If yes can you give me a brief about your implementation?

Thank you, Anmol Jain AI, Postgrad

On Sat, Oct 17, 2020 at 4:25 AM Zhedong Zheng notifications@github.com wrote:

Hi @anmol19005 https://github.com/anmol19005

Thanks a lot for your attention on our work. Sorry. We do not provide the cyclegan code. Actually, you can prepare data and train it easily via the official code ( https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix).

If you are interested in UNIT-like framework with generation and re-id, you may refer to my previous work DG-Net https://github.com/NVlabs/DG-Net

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/layumi/AICIty-reID-2020/issues/26#issuecomment-710687447, or unsubscribe https://github.com/notifications/unsubscribe-auth/AM7JMK4NGSQBNZBJD5TOT3TSLDFPBANCNFSM4Q6VUYGA .

Martafdez18 commented 3 years ago

Hi Zhedong,

I wanted to ask how do you train the Cycle-GAN. Do you use the synthetic and real data given by the challenge? However, you need real-synthetic vehicle pairs, which we don't have, right?

Thank you, Marta Fernández,

layumi commented 3 years ago

Hi @Martafdez18

Cycle-GAN do not require pairs, but two domain sets. In fact, you only need two folders to train cycle-gan for style changing.

anmol19005 commented 3 years ago

just go to cycle-gan repo and follow the instructions.

On Fri, Feb 12, 2021 at 4:46 PM Zhedong Zheng notifications@github.com wrote:

Hi @Martafdez18 https://github.com/Martafdez18

Cycle-GAN do not require pairs, but two domain sets. In fact, you only need two folders to train cycle-gan for style changing.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/layumi/AICIty-reID-2020/issues/26#issuecomment-778134756, or unsubscribe https://github.com/notifications/unsubscribe-auth/AM7JMK7VTJYLDZL5KKOMZ33S6UEYTANCNFSM4Q6VUYGA .