Валентин Сичкар

Design, Train & Test deep CNN for Image Classification: https://www.udemy.com/course/convolutional-neural-networks-for-image-classification

Content of the CNN Course
Section 1: Quick Win - Convolution inside-out
Section 2: Assemble datasets
Section 3: Modify datasets
Section 4: Preprocess datasets
Section 5: Design deep CNNs
Section 6: Train & Test CNNs
Section 7: Practice Test
Section 8: Augment datasets

Example of Image Classification


Concept Map of the CNN Course


Join the Course now!
https://www.udemy.com/course/convolutional-neural-networks-for-image-classification

Model #1 is built on pure 'numpy' and reached 0.98 Validation Accuracy.
Description of implementation with pure 'numpy' is available on GitHub here
Code implemented on Keras is available on Kaggle here

Related Paper:
Sichkar V. N. Effect of various dimension convolutional layer filters on traffic sign classification accuracy. Scientific and Technical Journal of Information Technologies, Mechanics and Optics, 2019, vol. 19, no. 3, pp. DOI: 10.17586/2226-1494-2019-19-3-546-552 (Full-text available here ResearchGate.net/profile/Valentyn_Sichkar)

Related Course:
Training YOLO v3 for Objects Detection with Custom Data. Build your own detector by labelling, training and testing on image, video and in real time with camera. Available here: https://www.udemy.com/course/training-yolo-v3-for-objects-detection-with-custom-data/

Model #1 has following architecture:
Conv -> ReLU -> Pooling -> Affine -> ReLU -> Affine -> Softmax

Initial Parameters	Description
Weights Initialization	HE Normal
Weights Update Policy	Adam
Activation Function	ReLU
Regularization	L2
Pooling	Max
Loss Function	Cross-entropy, Softmax

Hyperparameters	Description
Filters for ConvNet Layer	32
Size of the Filters	height=width=7
Params for ConvNet Layer	stride=1, pad=3
Params for Pooling Layer	stride=2, height=width=2
Dimension of Hidden Affine Layer	500 neurons
Regularization	1e-3

Training process of Model #1 with 18 000 iterations is shown on the figure below:



Initialized Filters and Trained Filters for ConvNet Layer is shown on the figure below:



Traffic Sign Classification with Model #1 in Real Time.
Left: original frame with detected Sign.
Upper Right: cut frame with detected Sign.
Lower Right: classified frame by ConvNet according to the detected Sign.