You will master the main types of CNN layers and identify larger portions of the images thanks to this 100% online program"

Convolutional Networks have established themselves as a versatile tool in the field of Computer Vision. Its importance lies in its ability to analyze, understand and process images or videos in an automated and efficient way. Among the diversity of its applications, it stands out its relevance in Biomedical Authentication by analyzing unique facial characteristics of a person and comparing them with a database to verify their identity. This is indispensable in aspects such as airport security or access control in buildings, among others.  

In this context, TECH is developing a Postgraduate diploma that will comprehensively address Deep Learning Applied to Computer Vision. The curriculum will delve into the use of Machine learning, given its importance to recognize patterns and perform specific analysis tasks. Likewise, the syllabus will address the whole cycle of creation of a Neural Network, paying careful attention to its learning and validation. On the other hand, students will learn the most advanced strategies for Object Detection and Tracking. In line with this, they will implement cutting-edge evaluation metrics, including the Intersection Over Union or Confidence Score. 

On the other hand, to strengthen the mastery of the contents, this university program applies the revolutionary Relearning system. TECH is a pioneer in the use of this teaching model, which promotes the assimilation of complex concepts through their natural and progressive reiteration. In this way, students do not have to resort to complex techniques such as traditional memorization. In this line, the program also uses materials in various formats such as infographics, interactive summaries or explanatory videos. All this in a convenient 100% online mode, which allows students to adjust their schedules according to their responsibilities and personal circumstances. 

Delve deeper into the Evaluation Metrics of Tracking Algorithms thanks to TECH, the world's best digital university according to Forbes"  

This Postgraduate diploma in Deep Learning Applied to Computer Vision contains the most complete and up-to-date program on the market. The most important features include:

• The development of case studies presented by experts in Deep Learning, computer science and computer vision
• The graphic, schematic, and practical contents with which they are created, provide scientific and practical information on the disciplines that are essential for professional practice
• Practical exercises where the self-assessment process can be carried out to improve learning
• Its special emphasis on innovative methodologies 
• Theoretical lessons, questions to the expert, debate forums on controversial topics, and individual reflection assignments
• Content that is accessible from any fixed or portable device with an Internet connection

Do you want to become a Machine Learning expert? Achieve it in only 6 months with this innovative program"  

The program’s teaching staff includes professionals from the industry who contribute their work experience to this program, as well as renowned specialists from leading societies and prestigious universities. 

The multimedia content, developed with the latest educational technology, will provide the professional with situated and contextual learning, i.e., a simulated environment that will provide immersive education programmed to learn in real situations. 

This program is designed around Problem-Based Learning, whereby the professional must try to solve the different professional practice situations that arise during the academic year For this purpose, the students will be assisted by an innovative interactive video system created by renowned and experienced experts.  

Update your knowledge in Object Detection through innovative multimedia content"

Forget about memorizing! With the Relearning system you will integrate the concepts in a natural and progressive way"

This study plan consists of 3 complete modules, designed by true specialists in Artificial Intelligence. Therefore, the didactic materials will offer the latest innovations in Neural Network evaluation metrics, types of CNN layers and training with regularization. In addition, students will acquire new skills to effectively handle the most advanced tools in object detection. The program will include the analysis of real cases and resolution of complex situations in simulated learning environments.” Graduates will be prepared to overcome any challenge that may arise during their activities. 

A specialist syllabus and top-level teaching materials are the key to a successful career"

Module 1. Deep Learning   

1.1. Artificial Intelligence

1.1.1. Machine Learning
1.1.2. Deep Learning
1.1.3. The Explosion of Deep Learning Why Now

1.2. Neural Networks

1.2.1. The Neural Network
1.2.2. Uses of Neural Networks
1.2.3. Linear Regression and Perceptron
1.2.4. Forward Propagation
1.2.5. Backpropagation
1.2.6. Feature Vectors

1.3. Loss Functions

1.3.1. Loss Functions
1.3.2. Types of Loss Functions
1.3.3. Choice of Loss Functions

1.4. Activation Functions

1.4.1. Activation Function
1.4.2. Linear Functions
1.4.3. Non-Linear Functions
1.4.4. Output vs. Hidden Layer Activation Functions

1.5. Regularization and Normalization

1.5.1. Regularization and Normalization
1.5.2. Overfitting and Data Augmentation
1.5.3. Regularization Methods: L1, L2 and Dropout
1.5.4. Normalization Methods: Batch, Weight, Layer

1.6. Optimization

1.6.1. Gradient Descent
1.6.2. Stochastic Gradient Descent
1.6.3. Mini Batch Gradient Descent
1.6.4. Momentum
1.6.5. Adam

1.7. Hyperparameter Tuning and Weights

1.7.1. Hyperparameters
1.7.2. Batch Size vs Learning Rate vs Step Decay
1.7.3. Weights

1.8. Evaluation Metrics of a Neural Network

1.8.1. Accuracy
1.8.2. Dice Coefficient
1.8.3. Sensitivity vs Specificity / Recall vs precision
1.8.4. ROC Curve (AUC)
1.8.5. F1-Score
1.8.6. Matrix Confusion
1.8.7. Cross-Validation

1.9. Frameworks and Hardware

1.9.1. Tensor Flow
1.9.2. Pytorch
1.9.3. Caffe
1.9.4. Keras
1.9.5. Hardware for the Training Phase

1.10. Creation of a Neural Network– Training and Validation

1.10.1. Dataset
1.10.2. Network Construction
1.10.3. Education
1.10.4. Visualization of Results

Module 2. Convolutional Neural Networks and Image Classification

2.1. Convolutional Neural Networks 

2.1.1. Introduction
2.1.2. Convolution
2.1.3. CNN Building Blocks

2.2. Types of CNN Layers 

2.2.1. Convolutional 
2.2.2. Activation
2.2.3. Batch Normalization 
2.2.4. Polling
2.2.5. Fully Connected 

2.3. Metrics

2.3.1. Matrix Confusion 
2.3.2. Accuracy 
2.3.3. Precision 
2.3.4. Recall
2.3.5. F1 Score 
2.3.6. ROC Curve 
2.3.7. AUC

2.4. Main Architectures 

2.4.1. AlexNet
2.4.2. VGG 
2.4.3. Resnet 
2.4.4. GoogleLeNet

2.5. Image Classification 

2.5.1. Introduction
2.5.2. Analysis of Data 
2.5.3. Data Preparation 
2.5.4. Model Training 
2.5.5. Model Validation

2.6. Practical Considerations for CNN Training 

2.6.1. Optimizer Selection
2.6.2. Learning Rate Scheduler
2.6.3. Check Training Pipeline 
2.6.4. Training with Regularization

2.7. Best Practices in Deep Learning 

2.7.1. Transfer Learning
2.7.2. Fine Tuning
2.7.3. Data Augmentation

2.8. Statistical Data Evaluation 

2.8.1. Number of Datasets
2.8.2. Number of Labels 
2.8.3. Number of Images 
2.8.4. Data Balancing

2.9. Deployment

2.9.1. Saving and Loading Models
2.9.2. Onnx
2.9.3. Inference

2.10. Case Study: Image Classification

2.10.1. Data Analysis and Preparation
2.10.2. Testing the Training Pipeline
2.10.3. Model Training
2.10.4. Model Validation

Module 3. Object Detection

3.1. Object Detection and Tracking

3.1.1. Object Detection
3.1.2. Case Uses
3.1.3. Object Tracking
3.1.4. Case Uses
3.1.5. Occlusions, Rigid and Non-Rigid Poses

3.2. Assessment Metrics

3.2.1. IOU-Intersection Over Union
3.2.2. Confidence Score
3.2.3. Recall
3.2.4. Precision
3.2.5. Recall–Precision Curve
3.2.6. Mean Average Precision (mAP)

3.3. Traditional Methods

3.3.1. Sliding Window
3.3.2. Viola Detector
3.3.3. HOG
3.3.4. Non-Maximal Suppresion (NMS)

3.4. Datasets

3.4.1. Pascal VC
3.4.2. MS Coco
3.4.3. ImageNet (2014)
3.4.4. MOTAChallenge

3.5. Two Shot Object Detector

3.5.1. R-CNN
3.5.2. Fast R-CNN
3.5.3. Faster R-CNN
3.5.4. Mask R-CNN

3.6. Single Shot Object Detector

3.6.1. SSD
3.6.2. YOLO
3.6.3. RetinaNet
3.6.4. CenterNet
3.6.5. EfficientDet

3.7. Backbones

3.7.1. VGG
3.7.2. ResNet
3.7.3. Mobilenet
3.7.4. Shufflenet
3.7.5. Darknet

3.8. Object Tracking

3.8.1. Classical Approaches
3.8.2. Particulate Filters
3.8.3. Kalman
3.8.4. Sort Tracker
3.8.5. Deep Sort

3.9. Deployment

3.9.1. Computing Platform
3.9.2. Choice of Backbone
3.9.3. Choice of Framework
3.9.4. Model Optimization
3.9.5. Model Versioning

3.10. Study: People Detection and Tracking

3.10.1. Detection of People
3.10.2. Monitoring of People
3.10.3. Re-Identification
3.10.4. Counting People in Crowds

A training program that is characterized by its flexibility, freedom of schedules and 24-hour availability. Enroll now!”