Object detection as a concept in computer vision
-
Introduction and course outline (6:40)
-
What is object detection for computer vision? (2:43)
-
Object detection can be for multiple objects in the image (1:46)
-
Why deep learning for object detection? (2:02)
-
The two categories of neural networks used for object detection (3:49)
-
High level overview of Faster RCNN (11:49)
-
High level overview of SSD (10:35)
-
High level overview of YOLOv3 (12:48)
How to choose the right neural network for your object detection task
-
Introduction (1:57)
-
Speed and accuracy tradeoffs in object detection (reference article) (1:26)
-
Accuracy VS time (plot in article) (5:19)
-
Object detector accuracy with respect to feature extractor accuracy (4:54)
-
Accuracy stratified by object size (4:25)
-
How the image resolution affects your object detection model (4:38)
-
How YOLO compares to SSD and Faster RCNN (8:39)
-
Summary on how to choose the right model for your object detection task (9:50)
Software setup
-
Linux installation : How to install tensorflow 2 with GPU support (part 1) (9:09)
-
Linux installation : How to install tensorflow 2 with GPU support (part 2) (11:14)
-
Linux installation : How to install tensorflow 2 object detection API (7:07)
-
Windows installation : Installing miniconda (2:15)
-
Windows installation : Create virtual environment (1:58)
-
Windows installation : Installing tensorflow 2 object detection API (13:23)
-
Windows installation : Installing tensorflow with GPU support (23:01)
Data for object detection
-
Data preparation for object detection (4:12)
-
The dataset that we will use to build an object detection model (9:38)
-
Downloading and setting up our annotation tool : labelImg (11:24)
-
Annotating the dataset (12:25)
-
Transforming our xml files into one csv file (12:39)
-
Creating a labelmap for our dataset (4:04)
-
The tool that we will use to generate tfrecords (1:59)
-
Generating tfrecords (7:09)
Training an object detection model using Tensorflow 2 object detection API
-
Overview of the steps needed to build an object detector (2:02)
-
Transfer learning (4:49)
-
Downloading the pretrained model and getting its corresponding config file (8:38)
-
Preparing your config file (10:33)
-
Running the training and testing for experimentation (18:38)
-
Running Tensorboard to analyse the development of the loss and precision (9:35)
-
Settings for training and evaluating a Faster RCNN model on your local machine (4:11)
-
Training and evaluating an SSD based object detection model (9:58)
-
Running the training for SSD object detector (10:05)
-
Other important settings in your config file : anchor boxes (10:52)
-
Other important settings in your config file : data augmentation (8:21)
Training object detection API models using Google Cloud AI Platform
-
What is cloud computing and what is AI Platform? (optional) (5:48)
-
Creating a Google Cloud account (5:28)
-
Downloading Google Cloud SDK (5:50)
-
Creating a google bucket and uploading data to it (8:32)
-
Preparing our config file for training on google cloud (7:51)
-
Running the training using Faster RCNN model (22:57)
-
Running the evaluation during the training (8:16)
-
Analyzing the results after the training of Faster RCNN model is finished (16:51)
-
Possible things to do to improve our model performance (2:52)
-
Downloading the trained model and exporting the frozen model from checkpoints (24:51)
-
Running the frozen model on new examples locally (20:03)
-
Run training and evaluation using SSD model (17:10)
-
Analyzing the results after the training of SSD model is finished (8:29)
YOLO v3 for object detection
-
The new setup for training YOLOv3 model (repository) (4:08)
-
Installing the necessary requirements to run the code (2:34)
-
How will the dataset change to account for YOLO neural network (4:49)
-
Preparing the dataset : problem with our current dataset (2:40)
-
Preparing the dataset : getting the data from the original source (5:38)
-
Preparing the dataset : transforming our previous dataset to the right format (17:01)
-
Preparing the dataset : adding classes names (1:38)
-
Preparing the dataset : exploring and changing the config file (16:08)
-
Training YOLO v3 based object detection model (3:29)
-
Analyzing the results of the training (8:40)
-
Evaluating our YOLOv3 trained model using new images (7:28)
-
Computing the mean average precision of our trained model (6:21)
-
Other quantitative results (3:03)
-
Saving our model as a SavedModel format for production deployment (6:33)
-
Using our trained model to make predictions on new images (16:57)
