Agenda
Deep Learning
-
Feed Forward Neural Network
-
Different Optimizer
-
Different FFN - Archt
-
Hidden Layers
-
How inforamtion is flowing from one neuron to another neuron
-
Different Activation Function
-
Different Loss Function
-
Back - Propagation
-
Building Feed Forward Neural Net
-
Assignmnet on FFN
-
Evaluation of FFNN
CNN
-
Archit of CNN
-
Padding
-
Strides
-
Conv layer, Max Polling,
-
Maths in Conv
-
Diff conv matrix to choose
-
Diff max pooling sizes to choose
-
Different filter matrix in Conv
-
2-d image features are extracted
-
3-d feature extracted
-
Which activation to use in different layers of CNN
-
What is FC - Fully connected layer
-
VGGNEt, Stylnet, Lenet
-
Using double conv layer
-
How we see an output in CNN, Argmax
-
Assigmnet
-
Emotion Detection - Project 1
-
Project - 2 - Different Garbage classification
RNN
-
Why RNN
-
Purpose of Tanh in RNN
-
Various Gates in RNN
-
Limitation of RNN
-
Build a model on RNN Network, songs or we use voices, NLP
-
LSTM's
-
Arch of LSTM Cell
-
Gates inside LSTM
-
Building a network with LSTM for Sequential Prediction - Voice Prediction
-
Bidirectional LSTM's
-
Try to submit a document - How RNNS can be used for Self-Driving Car
-
Project on LSTM - NLP /Voice
Autoencoders
-
DeConv Autoencoders
-
Sparse Autoencoders
-
Encoding and Decoding
-
Autoencoders for DR
-
Project - Blurred images and need to convert them into HD Images
Genetic Algorithm
-
Gene
-
Parent
-
Child
-
Mutation
-
Genetic Algorithm
-
Project - 1: Improving the model accuracy of an algorithm - FFNN by using different child's of GA
Markov Chains
-
Stochastic Mechanism
-
Probab of Events
-
Why Markov Chains
-
Limiataion of Markov Chain
-
Project : 1 NLP Text Mining
HMM-Hidden Markov Chains
-
Why HMM?
-
Prupose of HMM
-
Predicting the current event based on it's subevent
-
Predicting under closed state
-
Study: How Google Search Engine is working based on HMM
-
Study2 : How HMM Can be used for Self - Driving cars in Path planning & Localization
-
Project:1 HMM
-
Tensorflow Deployement
-
Tensorflow Parallel processing
-
Tensorflow Weight Saving
-
Weights importing
-
Tensorboard
Job Description
-
Building DNNs for Object detection, object classification and related problems
-
Working on real-time data and validate and improve algorithms
-
Experience on Computer Vision.
-
Experience on tools like Keras/Tensorflow/Caffe would be plus
-
Image segmentation/ Video Stabilisation/ Denoising of Image / General Object Detection
-
Experience on CNNs, RNNs and LSTM are must.
-
Experience on NLP with Deep Learning is required.
Job Requirements
-
Demonstrated ability to carry out independent research and lead projects
-
Experience with computer vision, machine learning, DNNs, and Numerical Optimization
-
Experience with big data, dataflow, data analytics, cloud storage and associated software tools.
-
Possess strong oral and written language skills
-
Present results for internal management and at external venues such as conferences
-
Experience on Tensorflow/Caffe
Learning Path
Level -1: Introduction to Artificial Intelligence
Various Branches in AI, How AI going to change the way we live. Detailed view on various applications of AI & Deep Learning
1. What is Text to Speech, Machine Translation, Robotics, ANNs
2. Understanding on sensors, Raspberry pi and build process.
Level -2: ANN Structure, Building Feed Forward
How ANNs are different from biological neurons. What is meant by Activation function, learning rate, batch size, epoch and loss functions. Comparing Human learning vs Deep Learning
1. Learning about Gradients
2. BackPropagation and Forward Propagation
Level -3: RNNs for Sequential Tasking; Feel the joy of coding
Deep understanding on building RNNs with the help of Tensorflow and Keras. Importance of LSTM building RNNs with LSTM cells
1. Work on Voice Recognition - RNN Network
2. Construct your own sequential tasking model with Tensorflow or Keras
Level -4: Computer Vision View - CNNs
In computer vision we recognise various human poses in like;
1. Body pose 2. Head Pose 3. Pupil Diameter 4. Face Detection
5. Face Classification 6. Drowsiness 7. Cognitive Load
1. We Implement CNN's to recognise various poses
2. Know about Strides, padding, conv layer, max pooling e.t.c.
3. Build CNNs on Localised Object Detection
Level -5: Blurred Image to HD Image - Autoencoders
Autoencoders gonna convert noised image into denoised and denoised to noised, Lets make build an AI system which can convert blurred image to hd image
Course Description
Deep Learning Engineers with Tensorflow is the course trending all the top MNC's are looking. Deep Learning is utilized broadly in the fields of image recognition, Natural Language Processing (NLP), self-driving cars, and video grouping.
Deep Learning has been started by top companies like Tesla, Audi, Benz, Ford and Google they are involved aggressively in building self-driving cars with the Deep Learning. At present there are very few companies started with Deep Learning Implementation but need for them are much more in future. Pay Scale of a Deep Learning engineer from 38 Lakh/Annum to 50 Lakh/Annum based on Experience.
Deep Learning Engineer is a blend of Python Programming, Computer Vision, Robotics, Object Detection e.t.c. Learn Deep Learning with various case studies and driven on various tools to make you best and bold on all topics of Deep Learning
Hiring Partners
Shyam
Data Science with R enhanced my knowledge and opened up various jobs
Abhilash
Excellent coding support from BEPEC team
Rupesh
Any person can learn R Programming such a wonderful support
Contact Us
Confused to choose between courses? Our counsellors are available to clarify your doubts and to lead you
Email:
Contact Number: 9121516181/9542548626
Visit our Office : Bangalore | Pune | Vijayawada | Mumbai | Hyderabad | Canada | United States