Page 358 - AI Ver 3.0 class 10_Flipbook
P. 358
Task 21 st Century #Information Literacy
Skills
CBSE Handbook
Convolutional Neural Network
Convolutional Rectified linear Pooling Fully Connected
Layer unit (ReLU) Layer Layer
10
1 1 1 0 0 8 Max [1,1,5,6]=6
x1 x0 x1
0 1 1 1 0 4 6 Car 70 %
x0 x1 x0
4
0 0 1 1 1 x Max pool with
x1 x0 x1 1 1 2 4 2×2 filters and Truck 20 %
2
0 0 1 1 0 5 6 7 8 stride 2 6 8
3 2 1 0 3 4
0 1 1 0 0 –10 –5 5 10 1 2 3 4 Bicycle 10 %
Convolved Output = Max(zero, Input)
Image Feature y
Rectified Features Map
Reduce size, improve feature, give probability value
Write the whole process of how a CNN works on the basis of the above diagram.
Testing CNN
To test a CNN, you typically need to perform the following steps:
1. Data Preparation: Collect a dataset suitable for your specific task, such as image classification or object
detection. Split the dataset into training, validation, and test sets.
2. Model Selection: Choose a CNN architecture suitable for your task, such as VGG, ResNet, or Inception. You may
also consider using pre-trained models to save time and resources.
3. Model Training: Train the CNN on the training dataset using an appropriate loss function and optimisation
algorithm. Monitor the model's performance on the validation set during training to avoid overfitting.
4. Evaluation: After training, evaluate the CNN's performance on the test dataset to assess its generalisation
ability. The common evaluation metrics for image classification tasks include accuracy, precision, recall and F1-
score.
356 Touchpad Artificial Intelligence (Ver. 3.0)-X

