Update README.md

README.md

@@ -4,7 +4,7 @@ This repository hosts the code and data lists for our two learning-based eXplain
 - This repository contains the code for training L-CAM-Fm and L-CAM-Img, using VGG-16 or ResNet-50 as the pre-trained backbone network along with the Attention Mechanism and our selected loss function. There is also code to train the above networks with the conventional cross-entropy loss. The models files with the model architecture are named as following:
 First the name of the backbone (ResNet50 or VGG-16) and then the method's name (L-CAM-Fm or L-CAM-Img). If the model uses the cross-entropy loss (instead of our proposed loss function) there is also an A character at the end of the name, e.g. ResNet50_L_CAM_ImgA.py. There is also a variation L-CAM-Img with VGG-16 backbone where the AM's input is the 7×7 FMs after the last max pooling layer of VGG-16, in contrast to all the others models that use the last convolutional layer of the backbone. This model is named VGG16_7x7_L_CAM_Img.py. 
 - Instead of training, the user can also download the pre-trained models for L-CAM-Fm and L-CAM-Img (again using VGG-16 or ResNet-50 as the backbone network along with the Attention Mechanism and our selected loss function [here](https://drive.google.com/drive/folders/1QiwB3iEobEPnSB9NRSsmDaUAuBMiPdz2?usp=sharing). The pre-trained models are named in the same way as the models files (as explained in the previous paragraph). 
-- There is also code for evaluating our method according to two widely used evaluation metrics for DCNN explainability, Increase in Confidence (IC) and Average Drop (AD).
+- There is also code for evaluating our method according to two widely used evaluation metrics for DCNN explainability, Increase in Confidence (IC) and Average Drop (AD). In the same script, Top-1 and Top-5 accuracy is as well calculated.
 - Furthermore, there is the code to evaluate the methods that are used in our paper for comparisons with L-CAM-Fm and L-CAM-Img.
 - In [L-CAM/datalist/ILSVRC](https://github.com/bmezaris/L-CAM/tree/main/datalist/ILSVRC) there are text files with annotations for training VGG-16 and ResNet-50 (VGG-16_train.txt, ResNet50_train.txt) and text files with annotations for 2000 randomly selected images to be used at the evaluation stage (Evaluation_2000.txt) for the L-CAM methods.
 - The ImageNet1K dataset images should be downloaded by the user manually.