Deep learning for object detection using Tensorflow 2

Name: Deep learning for object detection using Tensorflow 2
Rating: 4.5 (321 reviews)

Understand, train and evaluate Faster RCNN, SSD and YOLO v3 models using Tensorflow 2 and Google AI Platform

Created byNour Islam Mokhtari

Last updated 4/2023

English

What you'll learn

You will learn how Faster RCNN deep neural network works
You will learn how SSD deep neural network works
You will learn how YOLO deep neural network works
You will learn how to use Tensorflow 2 object detection API
You will learn how to train and evaluate deep neural networks for object detection such as Faster RCNN, SSD and YOLOv3 using your own custom data
You will learn how to "freeze" your model to get a final model that is ready for production
You will learn how to use your "frozen" model to make predictions on a set of new images using openCV and Tensorflow 2
You will learn how to use Google Cloud AI platform in order to train your object detection models on powerful cloud GPUs
You will learn how to use Tensorboard to visualize the development of the loss function and the mean average precision of your model
You will learn how to change different parameters in order to improve your model's performance

Course content

7 sections • 72 lectures • 9h 51m total length

Introduction and course content6:40
The lecture will outline the course content.
What is object detection for computer vision?2:43
In this lecture, we will define what object detection is in the context of computer vision.
Object detection can be for multiple objects in the image1:46
In this lecture, we will see how object detection can be used for detecting multiple objects of different categories inside the image.
Why deep learning for object detection?2:02
In this lecture, we will outline the reasons why we use deep learning for object detection instead of traditional image processing techniques.
The 2 categories of neural networks used for object detection3:49
In this lecture, we will list the 2 main classes of neural networks used in object detection.
High level overview of Faster R CNN11:50
In this lecture, I will show you how Faster RCNN network works from a high level point of view.
High level overview of SSD10:35
In this lecture, I will show you how SSD (Single Shot multibox Detector) network works from a high level point of view.
High level overview of YOLO v312:48
In this lecture, I will give you a high level overview to help you understand how YOLO algorithm works.

Introduction1:57
In this lecture I will explain to you what are the things that could affect your decision when tackling an object detection task.
Speed and accuracy tradeoffs in object detection (reference article)1:26
In this lecture, I will introduce you to one of the reference articles in the field of object detection in computer vision.
Accuracy VS inference time5:19
In this lecture, we will take a look at a plot that gives us an idea about how accuracy and inference time are affected, depending on our choice of the object detection model.
Object detector accuracy with respect to feature extractor accuracy4:54
In this lecture, we will see how the accuracy of the backbone (vgg16, inceptionV2, ...) affects the accuracy of the overall object detection model.
Accuracy stratified by object size4:25
By the end of this lecture, you will learn how the size of the object that you are trying to detect will affect the accuracy of your object detection model.
How the image resolution affects your object detection model4:38
In this lecture, you will learn about how the image resolution (width and height of the image) affect your model.
How YOLO compares to SSD and Faster R-CNN8:39
In this lecture, you learn how YOLO detector compares to SSD and Faster RCNN in terms of accuracy and time of inference.
Summary on how to choose the right model for your object detection task9:50
By the end of this lecture, you will learn how to choose your object detection model based on your pre-defined constraints.

Linux installation : How to install tensorflow with GPU support (part 1)9:09
In this lecture you learn how to install Tensorflow 2 with GPU support.
Linux installation : How to install tensorflow with GPU support (part 2)11:14
Continuing from the last lecture, you will continue installing all the necessary parts for Tensorflow 2 with GPU support.
Linux installation : How to install Tensorflow 2 object detection API7:07
In this lecture, you will learn how to install Tensorflow 2 object detection API.
Windows installation : Installing miniconda2:15
This lecture will show you how to install miniconda on your Windows machine.
Windows installation : Create virtual environment1:58
This video will show you how to create virtual envrironments on your windows machine using miniconda.
Windows installation : Installing tensorflow 2 object detection API13:23
In this lecture you will learn how to install Tensorflow 2 object detection API on your Windows machine
Windows installation : Installing tensorflow with GPU support23:01
In this video you will learn how to configure your Windows system so that you can run the training on the GPU using Tensorflow 2 and Cuda.

Data preparation for object detection4:12
In this lecture you will learn about data preparation for deep learning based object detection.
The dataset that you will use to build your object detection model9:38
In this video, I will show where to get the dataset that we will be using throughout the course to build our object detectors. The dataset contains images of people wearing masks and others who are not wearing masks. In the context of Covid19 this could be a very useful application!
Downloading and setting up our annotation tool : labelImg11:24
In this lecture, I will show you where to get and how to setup the tool (labelImg) that we will use to annotate our dataset.
Annotating your dataset12:25
In this lecture, we will go through some examples on how to annotate our dataset using labelImg.
Transforming our xml files into one csv file12:39
In this lecture you will learn how to transform you XML files into a CSV file.
Creating a labelmap file for our dataset4:04
In this video you will learn how to create a label map for your dataset.
The tool that we will use to generate tfrecords1:59
In this lecture, you will learn about the tool that we will use to transform our dataset into tfrecords format.
Generating tfrecords7:09
In this video, you will use the tool that we introduced earlier to generate tfrecords.

Overview of the steps needed to build an object detector2:02
In this lecture, you will learn about the different steps needed to build an object detection model using Tensorflow 2 object detection API.
Transfer learning4:49
This video will explain to you the concept of transfer learning in the context of object detection.
Downloading the pretrained model and getting its corresponding config file8:38
Preparing your config file10:33
This video will show you how to prepare your config file for training by changing the right parameters.
Running the training and testing for experimentation18:38
In this video, you will learn how to run the training and evaluation of your object detection task using scripts from the Tensorflow 2 object detection API.
Running Tensorboard to analyse the development of the loss and average precision9:35
This lecture will show you how to use Tensorboard to visualise the development of your training and evaluation.
Settings for training and evaluating a Faster RCNN model on your local machine4:11
In this lecture, you will learn how to create the right settings to train and evaluate a Faster RCNN based object detection model using a custom dataset.
Training and evaluating an SSD based object detection model
Solution : Training and evaluating an SSD based object detection model9:58
This lecture is for presenting the solution to the previous assignment. It will show you how to prepare things for training and evaluating a SSD based object detection model using a custom dataset.
Running the training for SSD object detector10:05
This lecture will show you how to run the training and evaluation of SSD based object detection model using a custom dataset.
Other important settings in your config file : anchor boxes10:52
This video will show you some of the important settings that you should know about and how to tune them for your specific object detection task. Specifically, you learn about tuning anchor boxes.
Other important settings in your config file : data augmentation8:21
This video will show you some of the important settings that you should know about and how to tune them for your specific object detection task. Specifically, you learn about tuning data augmentation options.

What is cloud computing and what is AI-Platform? (optional)5:48
This lecture will give you an overview of what cloud computing is.
Creating a Google Cloud account and creating your first project5:28
This video will show you how to create a google cloud account so that you can run your training and evaluation on the cloud.
Downloading Google Cloud SDK5:50
This video will show you where to download the SDK for google cloud platform.
Creating a google bucket and uploading data to it8:32
This lecture will show you how to create a google cloud bucket to store your data on google cloud platform just like you would do on your local drive.
Preparing our config file for training on google cloud7:51
In this video, I will show you how to modify your config file in order to account for the new setup on the cloud. You will change the paths to your data to point to your google bucket instead of your local drive.
Running the training using Faster RCNN model22:57
In this lecture, you will learn how to run the training of Faster RCNN model with you custom data on Google AI Platform.
Running the evaluation during the training8:16
In this video, you will learn how to run the evaluation of your Faster RCNN model during the training.
Analyzing the results after the training of Faster RCNN model is finished16:51
In this lecture, you will learn how to analyze the progress of your training and evaluation using Tensorboard and cloud shell on Google Cloud Platform.
Possible things to do to improve our model performance2:52
In this lecture, I will show you how to modify some parameters to improve our model performance.
Downloading the trained model and exporting the frozen model from checkpoints24:51
In this lecture, I will show you how to download your trained model and then "freeze" it to get a model that is ready for deployment.
Running the frozen model on new examples locally20:03
In this lecture, you will learn how to use your frozen model in order make predictions on new images.
Run a training and evaluation using SSD model
Solution : Run a training and evaluation using SSD model17:10
This video will show you the solution to the previous assignment. You will understand how to run the training and evaluation of a SSD based object detection model using Google AI Platform.
Analyzing the results after the training of SSD model is finished8:29
This lecture will show you how you can analyse your training and evaluation using Tensorboard and cloud shell on Google Cloud Platform.

The new setup for training YOLOv3 model4:08
In this video I will be showing you the new setup for training YOLO v3 deep learning model. Tensorflow 2 object detection API doesn't have support for any of the YOLO models so we need a new code base to train and evaluate this neural network.
Installing the necessary requirements to run the code2:34
In this video, I will be showing you how to install all the necessary requirements to be able to run the code for training and evaluation.
How will the dataset change to account for YOLO neural network4:49
In this lecture, I will be showing you the difference between the dataset that we used for training on Tensorflow 2 object detection API and the dataset that we will use to train YOLOv3.
Preparing the dataset : problem with our current dataset2:40
In this lecture, I will be showing you a problem with our previous dataset that prevents us from using it as it is.
Preparing the dataset : getting the data from the original source5:38
In this video, I will show you a possible source to get the data necessary for training YOLO v3 neural network.
Preparing the dataset : transforming our previous dataset to the right format17:01
In this lecture, I will show you how to transform our previous dataset into the right format for training YOLO v3 with Tensorflow 2.
Preparing the dataset : adding classes names1:38
In this video, I will show you how to add classes names to a text file, which is a necessary step before training our model.
Preparing the dataset : exploring and changing the config file16:08
In this lecture, I will show you the important parameters that you need to set when training YOLOv3 model. We will do all of his inside a config file.
Training YOLO v3 based object detection model3:29
In this lecture, we will start training our YOLO v3 model.
Analyzing the results of the training8:40
In this video, we will analyze the results of the training process.
Evaluating our YOLOv3 trained model using new images7:28
In this video, I will show you how to evaluate your trained YOLOv3 model.
Computing the mean average precision of our trained model6:21
In this lecture, I will show you how to compute mean average precision (mAP) metric to evaluate your model in a quantitative.
Other quantitative results3:04
In this lecture, we will explore some other quantitative metrics to understand our model performance.
Saving our model as a SavedModel format for production deployment6:33
In this video, I will show you how to export your SavedModel so that you can use your model in production.
Using our trained model to make predictions on new images16:57
In this lecture, we will use our final trained model to make predictions on new images.
Steps needed to run the training on google AI platform6:45
In this final lecture, I will show you the necessary steps to run the training of your YOLO v3 model on google AI platform. I will also guide you to the necessary resources to help you do so.
Bonus materials0:07

Requirements

You need to have a basic level of Python (if you know what classes and functions are then you are good to go!)
You need to have a basic understanding of what Tensorflow is.
You don't need any prior understanding of what object detection is, this is the mission of the course!

Description

This course is designed to make you proficient in training and evaluating deep learning based object detection models. Specifically, you will learn about Faster R-CNN, SSD and YOLO models.

For each of these models, you will first learn about how they function from a high level perspective. This will help you build the intuition about how they work.

After this, you will learn how to leverage the power of Tensorflow 2 to train and evaluate these models on your local machine.

Finally, you will learn how to leverage the power of cloud computing to improve your training process. For this last part, you will learn how to use Google Cloud AI Platform in order to train and evaluate your models on powerful GPUs offered by google.

I designed this course to help you become proficient in training and evaluating object detection models. This is done by helping you in several different ways, including :

Building the necessary intuition that will help you answer most questions about object detection using deep learning, which is a very common topic in interviews for positions in the fields of computer vision and deep learning.
By teaching you how to create your own models using your own custom dataset. This will enable you to create some powerful AI solutions.
By teaching you how to leverage the power of Google Cloud AI Platform in order to push your model's performance by having access to powerful GPUs.

Who this course is for:

AI enthusiasts
Data scientists
Computer vision and machine learning students
software developers
Entrepreneurs

Deep learning for object detection using Tensorflow 2

What you'll learn

Explore related topics

Course content

Object detection as a concept in computer vision8 lectures • 52min

How to choose the right neural network for your object detection task8 lectures • 41min

Software setup7 lectures • 1hr 8min

Data for object detection8 lectures • 1hr 4min

Training an object detection model on your local machine11 lectures • 1hr 38min

Training object detection API models using Google Cloud AI Platform13 lectures • 2hr 35min

YOLO v3 for object detection17 lectures • 1hr 54min

Requirements

Description

Who this course is for: