dougsouza / efficient-t2i Goto Github PK
View Code? Open in Web Editor NEWOfficial implementation of the paper Efficient Neural Architecture for Text-to-Image Synthesis.
Home Page: https://arxiv.org/abs/2004.11437
Official implementation of the paper Efficient Neural Architecture for Text-to-Image Synthesis.
Home Page: https://arxiv.org/abs/2004.11437
Hi.
Thank you for opening codes of efficiet-t2i. I've been interested in this since the paper was uploaded to arxiv. I would like to try on my dataset. I predict this file(CUB_200_2011/birds/train/attn_embeddings.npy
) was made by the AttnGAN's prepare_DAMSM.py, is it correct?
I look forward to hearing from you.
How do you evaluate the IS on oxford-102? Did you use the inception_finetuned_models provided by stackGAN? But eval.py looks only write for CUB-birds.
Pretrained birds model link in scripts/download_pretrained_birds_model.sh
gives an access denied response:
Access denied with the following error:
Cannot retrieve the public link of the file. You may need to change the permission to 'Anyone with the link', or have had many accesses.
You may still be able to access the file from the browser:
https://drive.google.com/uc?id=1YLqAkHuyPWof64amelOie2t2NwRdmsWk
Opening the link in the browser gives a 403
error.
There is no download_flowers.sh in scripts, could you upload it?
Hi.
I'm Haruka.
I would like to ask about attn_embeddings.npy
again. I checked that the shape of CUB's attn_embeddings.npy
is (8855, 10, 256). I thiught that shpe is (data_length, seq_size, nef). And, I tried to generate the embeddings from my dataset, but I got the shape (data_length, 14, 17, 17).
I created the following code by appropriating the following from AttnGAN and created it, is it wrong? I treat the attn
is a attn_embeddings.
list_attn_map = []
list_class_info = []
list_keys = []
with torch.no_grad():
for step, data in enumerate(dataloader, 0):
real_imgs, captions, cap_lens, \
class_ids, keys = prepare_data(data)
# words_emb: batch_size x nef x seq_len
# sent_emb: batch_size x nef
words_features, sent_code = cnn_model(real_imgs[-1])
# nef, att_sze = words_features.size(1), words_features.size(2)
# _words_features = words_features.view(batch_size, nef, -1)
hidden = rnn_model.init_hidden(batch_size)
words_emb, sent_emb = rnn_model(captions, cap_lens, hidden)
w_loss0, w_loss1, attn = words_loss(words_features, words_emb, labels,
cap_lens, class_ids, batch_size)
for j, (key, attn_map, cap_len, class_id) in enumerate(zip(keys, attn, cap_lens, class_ids)):
print( "[%d/%d] [%d/%d]\t%s" % (j+1, len(keys), step, len(dataloader), key) )
np_attn = attn_map.to('cpu').detach().numpy().copy()[0]
list_attn_map.append(np_attn)
list_class_info.append(class_id)
list_keys.append(key)
return np.array(list_attn_map), list_keys, list_class_info
Thank you.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.