Created
May 13, 2020 13:01
-
-
Save gauravbansal98/95d7b0255a8e6fb5bdcadf3ece952fdc to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
class CaptionModel(nn.Module): | |
def __init__(self, vocab_size): | |
super(CaptionModel, self).__init__() | |
# Input from the encoder decoder | |
self.feature_extractor = nn.Sequential() | |
self.feature_extractor.add_module('dropout', nn.Dropout(0.5)) | |
self.feature_extractor.add_module('FC', nn.Linear(in_features=4096, out_features=256)) | |
self.feature_extractor.add_module('activation', nn.ReLU()) | |
# Sequence Model | |
self.embeddings = nn.Embedding(num_embeddings=vocab_size, embedding_dim=256, padding_idx=0) | |
self.emb_drop = nn.Dropout(0.5) | |
self.lstm = nn.LSTM(input_size=256, hidden_size=256, batch_first=True, num_layers=1) | |
# Decode Model | |
self.decoder = nn.Sequential() | |
self.decoder.add_module('FC1', nn.Linear(in_features=256, out_features=256)) | |
self.decoder.add_module('activation', nn.ReLU()) | |
self.decoder.add_module('FC2', nn.Linear(in_features=256, out_features=vocab_size)) | |
def forward(self, image, caption): | |
image = self.feature_extractor(image) | |
embd = self.embeddings(caption) | |
embd_drop = self.emb_drop(embd) | |
outputs, (hidden, _) = self.lstm(embd_drop) | |
hidden = hidden.view(hidden.size(1), -1) | |
decoder_features = image + hidden | |
out = self.decoder(decoder_features) | |
return out |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment