search

tobran / DF-GAN

[CVPR2022 oral] A Simple and Effective Baseline for Text-to-Image Synthesis
Other
297 stars 67 forks source link

Abnormal Inception Score on the CUB dataset #8

Closed senmaoy closed 2 years ago

senmaoy commented 3 years ago

I got a result of mean: 4.16 std: 0.15 on the CUB dataset, which is much worse than 5.1. I did not change any code.

skabbit commented 3 years ago

What training parameters did you use?

senmaoy commented 3 years ago

What training parameters did you use?

TRAIN: NF: 32 # default 64 BATCH_SIZE: 24 MAX_EPOCH: 601 NET_G: '../test'

I also try NF:64 with batch_size 22. Then the IS increase to 4.46 std 0.15. I use 2925 images for evaluation. I notice that there is no truncation code implemented in the code.

lxdcsu commented 3 years ago

I got a result of mean: 4.41 std: 0.18 on the CUB dataset, which is much worse than 5.1.

PeleLiu777 commented 2 years ago

What training parameters did you use?

TRAIN: NF: 32 # default 64 BATCH_SIZE: 24 MAX_EPOCH: 601 NET_G: '../test'

I also try NF:64 with batch_size 22. Then the IS increase to 4.46 std 0.15. I use 2925 images for evaluation. I notice that there is no truncation code implemented in the code.

These methods each model generate 30,000 images for evaluation

lxdcsu commented 2 years ago

------------------ 原始邮件 ------------------ 发件人: "tobran/DF-GAN" @.>; 发送时间: 2021年11月9日(星期二) 晚上7:15 @.>; @.**@.>; 主题: Re: [tobran/DF-GAN] Abnormal Inception Score on the CUB dataset (#8)

My training parameters and the IS are the same as yours.

What training parameters did you use?

TRAIN: NF: 32 # default 64 BATCH_SIZE: 24 MAX_EPOCH: 601 NET_G: '../test'

I also try NF:64 with batch_size 22. Then the IS increase to 4.46 std 0.15. I use 2925 images for evaluation. I notice that there is no truncation code implemented in the code.

These methods each model generate 30,000 images for evaluation

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.