Skip to content

Latest commit

 

History

History
59 lines (38 loc) · 3.11 KB

README.md

File metadata and controls

59 lines (38 loc) · 3.11 KB

Convolution Neural Network - Image Question Answering

This is a python and keras implementation of the VIS+LSTM visual question answering model. This model is explained in the paper Exploring Models and Data for Image Question Answering. A second model is also implemented which is similar to the 2-VIS+BLSTM model from the paper mentioned above except that the LSTMs are not bidirectional. This model has two image feature inputs, at the start and the end of the sentence, with different learned linear transformations. We call it 2-VIS+LSTM.

Details about the dataset are explained at the VisualQA website.

Here is a summary of performance we obtained on both the models.

Model Epochs Batch Size Validation Accuracy
VIS+LSTM 10 200 53.27%
2-VIS+LSTM 10 200 54.01%

Requirements

  • Python 2.7
  • Numpy
  • Scipy (for loading pre-computed MS COCO features)
  • NLTK (for tokenizer)
  • Keras
  • Theano

Training

  • The basic usage is python train.py.

  • The model to train can be specified using the option -model. For example, to train the VIS+LSTM model enter python train.py -model=1. Similarly, the 2-VIS+LSTM model can be trained using python train.py -model=2. If no model is specified, model 1 is selected.

  • The batch size and the number of epochs can also be specified using the options -num_epochs and -batch_size. The default batch size and number of epochs are 200 and 25 respectively.

  • To train 2-VIS+LSTM with a batch size of 100 for 10 epochs, we would use: python train.py -model=2 -batch_size=100 -num_epochs=10.

Models

VIS+LSTM

2-VIS+LSTM

Prediction

  • Q&A can be performed on any image using the script question_answer.py.

  • The options -question and -image are used to specify the question and address of the image respectively. The model to use for the prediction can be specified using -model. By default, model 2 is selected.

  • An example of usage is: python question_answer.py -image="examples/COCO_val2014_000000000136.jpg" -question="Which animal is this?" -model=2

Here are some examples of predictions using the 2-VIS+LSTM model.

Image Question Top Answers (left to right)
Which animal is this? giraffe, cat, bear
Which vehicle is this? motorcycle, taxi, train
How many dishes are there? 5, 3, 2
What is in the bottle? water, beer, wine
Which sport is this? tennis, baseball, frisbee