Deep Learning for Computer Vision with Tensorflow 2

This course offers a complete review of the Deep Learning application in Computer Vision, especially on tasks in Image Classification and Object Detection.

The course was entirely written using Google Colaboratory(Colab) with Tensorflow 2.X, to help students who don't have a GPU card in their local system, however, you can follow the course easily if you have one.

We're going to study in detail the following concepts and algorithms:

• Image Fundamentals in Computer Vision,
• Load images in Generators with TensorFlow,
• Convolution Operation,
• Sparsity Connections and parameter sharing,
• Depthwise separable convolution,
• Padding,
• Conv2D layer with Tensorflow,
• Pooling layer,
• Fully connected layer,
• Batch Normalization,
• ReLU activation and other functions,
• Number of training parameters calculation,
• Image Augmentation, etc
• Different ConvNets architectures such as:
• LeNet5,
• AlexNet,
• VGG-16,
• ResNet,
• Inception,
• The latest state of art Vision Transformer (ViT)

Many practical applications using famous datasets and sources such as:

• COVID-19 on X-Ray images,
• CIFAR10,
• Fashion MNIST,
• BCCD,
• COCO dataset,
• Open Images Dataset V6 through Voxel FiftyOne,
• ROBOFLOW

In the Object Detection chapter, we'll learn the theory and the application behind the main object detection algorithms doing a journey from the beginnings to the latest state-of-the-art algorithms. You'll be able to use the main algorithms of object detection to develop practical applications.

Some of the content in this Chapter is the following:

• - Object detection milestones since the Selective Search algorithm,
• - Object detection metrics,
• Theoretical background for R-CNN, Fast R-CNN and Faster R-CNN.
• Detect blood cells using the Faster R-CNN application.
• Theoretical background for Single Shot Detector (SSD),
• Train your customs datasets using different models with TensorFlow Object Detection API.
• Blood Cells detection application.
• Object Detection on images and videos.
• YOLOv2 and YOLOv3 theory.
• Object detection from COCO dataset application using the YOLOv4 model.
• YOLOv4 theoretical class.
• Practical application for detecting Robots using a custom dataset (R2D2 and C3PO robots dataset) and YOLOv4 model.
• Practical application for License Plate recognition converting the plate images in raw text format (OCR) with Yolov4, OpenCV and ConvNets.
• Object detection with the latest state-of-the-art YOLOv7.
  You will find in this course a concise review of the theory with intuitive concepts of the algorithms, and you will be able to put into practice your knowledge with many practical examples using your own datasets.