computer vision using neural networks

2. Deep Learning
Machine Learning
Deep Learning
Input Feature extraction Classification Output
Input Feature extraction | Classification Output

3. Feed Forward Network
Feed forward neural network (FFNN) model topology with single hidden layer

4. Back Propogation

5. Activation Function

6. CNN

7. Transfer Learning

8. Convolution Layers

9. ⚫ Lenet
⚫ Alexnet
⚫ VGG
⚫ Inception
⚫ Resnet
Types of Architecture

10. Lenet

11. Alexnet

12. VGG

13. Object Detection

14. Word Embeddings
➢ Word embedding is a methodology to map words to a
corresponding vector.
• This process can also be called vectorization
• It can compute similiar words and text processing

15. CountVectorizer
CountVectorizer is used to transform a given text into a
vector on the basis of the frequency (count) of each word
that occurs in the entire text

16. TF-IDF

17. CountVectorizer
CountVectorizer is used to transform a given text into a vector
on the basis of the frequency (count) of each word that occurs in
the entire text

18. How 2 words are compared?
Cosine Distance/Similarity

19. Word2Vector
➢ Word2vector consist of two techniques
• CBOW model
• Skip Gram model

20. RNN vs LSTM

21. Cbow vs Skip-gram

22. CNN vs RNN
Recurrent neural network (RNN)
Recuring network that feeds the results
back into the network
The size of the input and the resulting
output may very (i.e, receives different
text and output translations – the
resulting sentences can have more of
fewer words)
Temporal / Sequential data (such as text
or video
Text translation natural language
processing, language translation, entity
extraction, conversational intelligence,
sentiment analysis, speech analysis
Convolutional neural network (CNN)
Feed-forward neural networks using
filters and pooling
The size of the input and the resulting
output are fixed (i.e., receives images of
fixed size and outputs them to the
appropriate category along with the
confidence level of its prediction)
Spatial data (such as images)
Image recognition and classification,
face detection, medical analysis, drug
discovery and image analysis
ARCHITECTURE
INPUT/OUTPUT
IDEAL USAGE
SCENARIO
USE CASES

23. Seq2Seq Model

24. Encoder Decoder

25. Bidirectional RNN