GithubHelp home page GithubHelp logo

leiloong / lisgan Goto Github PK

View Code? Open in Web Editor NEW

This project forked from lijin118/lisgan

0.0 1.0 0.0 25 KB

LisGAN, Leveraging the Invariant Side of Generative Zero-Shot Learning, CVPR 2019

Python 100.00%

lisgan's Introduction

LisGAN

LisGAN, Leveraging the Invariant Side of Generative Zero-Shot Learning, CVPR 2019, the pdf can be found Here

Just Run LisGAN.py and have fun!

If you find it is helpful, plese cite

@inproceedings {Li19Leveraging, 	
 title = {Leveraging the Invariant Side of Generative Zero-Shot Learning}, 	
 booktitle = {IEEE Computer Vision and Pattern Recognition (CVPR)}, 	
 year = {2019}, 	
 author = {Li, Jingjing and Jing, Mengmeng and Lu, Ke and Ding, Zhengming and Zhu, Lei and Huang, Zi} 
} 

Many Thanks!

Datasets can be downloaded Here

If you find that the results are prone to 0, please check that you run the code with pytorch 0.3.1 or change the accuracy calculation code from torch.sum(....) to torch.sum(....).float()

Here are some samples for your reference:

# zsl_cub
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1e-2 --proto_param2 0.001 --ratio 0.6  --manualSeed 3483 --val_every 1 --cls_weight 0.01 --preprocessing --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 70 --ngh 4096 --ndh 4096 --lr 0.0001 --classifier_lr 0.001 --lambda1 10 --critic_iter 5 --dataset CUB --batch_size 64 --nz 312 --attSize 312 --resSize 2048 --syn_num 300 --outname cub 

# zsl_flo
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1e-4 --proto_param2 0.001 --ratio 0.6  --manualSeed 806 --cls_weight 0.1 --syn_num 300 --preprocessing --val_every 1 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 97 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset FLO --batch_size 64 --nz 1024 --attSize 1024 --resSize 2048 --lr 0.0001 --outname flowers

# zsl_sun
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 3e-1 --proto_param2 3e-4 --ratio 0.5  --manualSeed 4115 --cls_weight 0.01 --val_every 1 --preprocessing --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 54 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset SUN --batch_size 64 --nz 102 --attSize 102 --resSize 2048 --lr 0.0002 --classifier_lr 0.0005 --syn_num 100 --outname sun

# zsl_awa
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 3e-2 --proto_param2 3e-5 --ratio 0.1  --manualSeed 9182 --cls_weight 0.01 --preprocessing --val_every 1 --lr 0.00001 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 30 --syn_num 300 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset AWA1 --batch_size 64 --nz 85 --attSize 85 --resSize 2048 --outname awa 

# zsl_apy
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1    --proto_param2 3e-5 --ratio 0.7  --manualSeed 9182 --cls_weight 0.01 --preprocessing --val_every 1 --lr 0.00001 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 40 --syn_num 300 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset APY --batch_size 64 --nz 64 --attSize 64 --resSize 2048 --outname apy

# gzsl_cub
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1e-2 --proto_param2 0.001 --ratio 0.2 --gzsl --manualSeed 3483 --val_every 1 --cls_weight 0.01 --preprocessing --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 56 --ngh 4096 --ndh 4096 --lr 0.0001 --classifier_lr 0.001 --lambda1 10 --critic_iter 5 --dataset CUB --nclass_all 200 --batch_size 64 --nz 312 --attSize 312 --resSize 2048 --syn_num 300 --outname cub

# gzsl_flo
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1e-1 --proto_param2 3e-2 --ratio 0.4 --gzsl --nclass_all 102 --manualSeed 806 --cls_weight 0.1 --syn_num 1200 --preprocessing --val_every 1 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 80 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset FLO --batch_size 64 --nz 1024 --attSize 1024 --resSize 2048 --lr 0.0001 --classifier_lr 0.001 --outname flowers

# gzsl_sun
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 3e-1 --proto_param2 3e-5 --ratio 0.1 --gzsl --manualSeed 4115 --cls_weight 0.01 --val_every 1 --preprocessing --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 40 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --dataset SUN --batch_size 64 --nz 102 --attSize 102 --resSize 2048 --lr 0.0002 --syn_num 400 --classifier_lr 0.001 --nclass_all 717 --outname sun 

# gzsl_awa
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 1e-3 --proto_param2 3e-5 --ratio 0.1 --gzsl --manualSeed 9182 --cls_weight 0.01 --preprocessing --val_every 1 --lr 0.00001 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 30 --syn_num 1800 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --nclass_all 50 --dataset AWA1 --batch_size 64 --nz 85 --attSize 85 --resSize 2048 --outname awa 

# gzsl_apy
CUDA_VISIBLE_DEVICES=0 python3 ../LisGAN.py --proto_param1 3e-1 --proto_param2 3e-4 --ratio 0.2 --gzsl --manualSeed 9182 --cls_weight 0.01 --preprocessing --val_every 1 --lr 0.00001 --cuda --image_embedding res101 --class_embedding att --netG_name MLP_G --netD_name MLP_CRITIC --nepoch 50 --syn_num 1800 --ngh 4096 --ndh 4096 --lambda1 10 --critic_iter 5 --nclass_all 32 --dataset APY --batch_size 64 --nz 64 --attSize 64 --resSize 2048 --outname apy 

lisgan's People

Contributors

lijin118 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.