README.md 473 Bytes
Newer Older
Arpit Lekhra's avatar
Arpit Lekhra committed
1
# Test task for Context embedding in Image Captioning
Arpit Lekhra's avatar
Arpit Lekhra committed
2
Data provided in test task is used.
Arpit Lekhra's avatar
Arpit Lekhra committed
3
Have used alexnet for encoder and lstm for decoder.
Arpit Lekhra's avatar
Arpit Lekhra committed
4
5
6
7
8
Have used Flickr8k.token.txt for making vocab.
Code was developed in Google colab, if you are running it in colab, no dependencies needed to download.
For running code, you have to handle paths of files where ever used.
final result: Loss: 0.0022, Perplexity: 1.0022.
Does not contain code for visualising output images as of now.