Dataset: CIFAR-10
Aim: Visualisation of deep learning model using GradCam.
Model: designed a deep learning model and train it for 10-class classification.
Results
a. The accuracy obtained on the test set.
b. Select 10 samples from each class which were correctly classified by the trained model. Apply GradCam on it and visualise most salient regions being used for prediction.
c. Select 5 samples from each class which were incorrectly classified by the trained model and apply GradCam to visualise most salient regions along with the predicted class and confidence.
Reference Article: https://towardsdatascience.com/demystifying-convolutional-neural-networks-using-gradcam-554a85dd4e48
Reference Paper: https://arxiv.org/pdf/1610.02391.pdf