-
LeNet-5
-
DenseNet169
-
ResNet152V2
-
Vgg16
- First, download the data file, load it, and
- Describe the data
- Clean the data
- Check the data for missing values or duplicates and carry out proper correction methods
- Visualize the data using proper visualization methods.
- Draw some of the images
- Carry out required correlation analysis
- Carry out any required preprocessing operations on the data
- Encode the labels
-
Modify hyperparameters to get to the best performance you can achieve.
-
Evaluate the model using 5-fold cross-validation.
-
In the report, provide a plot of accuracy improvement using the previously mentioned techniques. Also plot the convergence curve for LeNet-5.
-
Comment on why you think LeNet-5 further improves the accuracy if any at all. And if it doesn't, why not?
-
Note that this is an open-ended problem. So, there could be very good solutions.
-
Try to use other two CNN models (using transfer learning) and compare the results with the full trained LeNet-5.