Evaluating our image captioning deep learning model

Training a model and not evaluating its performance makes no sense at all. Hence, we will now be evaluating our deep learning model's performance on the test dataset, which has a total of 1,000 different images from the Flickr8K dataset. We start off by loading up the usual dependencies in case you don't already have them:

import pandas as pd 
import numpy as np 
import matplotlib.pyplot as plt 
 
pd.options.display.max_colwidth = 500 

%matplotlib inline 

Loading up data and models

The next steps include loading up the necessary data, model, and other assets from disk into memory. We first load up our test dataset and our trained deep learning models:

# load test dataset 
test_df = pd.read_csv('image_test_dataset.tsv', delimiter='\t') 
 
# load the models 
from keras.models import load_model 
 
model1 = load_model('ic_model_rmsprop_b256ep30.h5') 
model2 = load_model('ic_model_rmsprop_b256ep50.h5') 

We now need to load up necessary metadata...