YOLOv4 Object Detection Course

Learn to prepare and annotate image and video data for YOLOv4, using bounding boxes and other annotation methods, and apply semi-automated labeling to train a mask detector.

Learn to prepare data in YOLO format for image labeling by creating per-image text files named like the image, containing normalized class, center coordinates, and dimensions.

Discover free, easy to use image labeling tools for bounding boxes, including label Angi and others, with features like bulk upload, YOLO format export, and video tracking for rapid annotation.

Learn to build a dataset for yolov4 object detection by web-scraping with a chrome extension and Google image search, balancing classes: person with a mask and person without a mask.

Label objects in your image dataset with the LabelImg tool, draw bounding boxes, ensure every image has a label and remove irrelevant images, then convert annotations to yolo format.

practice labeling images in YOLOv4 format using the label IMG Tool, then report how many objects you labeled and what labels you gave them in the comments.

Transform video into frames, extract at 15 fps, and label with labelImg by drawing bounding boxes for faces with or without masks and for glasses to build a YOLOv4 dataset.

Use darklabel to label long videos efficiently by applying a semi-autonomous tracking workflow with yolo ground-truth annotations, interpolation, and keyboard shortcuts for rapid labeling.

Convert the video into image frames per second, label objects for two classes: mosque and without mosque, using the YOLO labeling workflow, and compare two labeling tools.

Create a dataset by collecting images and videos, edit with img or darktable, and annotate; define mosque and non-mosque classes, apply web scraping and data augmentation for YOLO labels.

Summarize data annotation for YOLOv4 object detection by labeling with binding boxes in yellow v4 format on images and videos, and outsourcing work to freelancers or third parties.

Learn to create custom datasets for yolov4 by extracting target classes from open images and converting annotations to V4 format for efficient detector training.

Learn to install and use the OOIDA Toolkit to download selective image classes from large datasets for object detection, enabling multi-class downloads with minimal effort.

Download images and annotations for specific classes from the open images dataset using a python tool with the --classes flag, using underscores for multiword names, and verify with the visualizer.

Practice using the Open Images dataset toolkit to download ten images per class into separate folders, and review commands, class naming rules, and dataset selection.

Convert downloaded csv annotations to YOLOv4 format by creating per-image text files with center-based bounding boxes, and verify results with labelImg.

Boost YOLOv4 training with rotational data augmentation, expanding images and annotations via a rotation script. Configure input, run the script, and verify augmented results in the rotational dataset.

Downloading labeled images with a toolkit, converting annotations to v4 format, and applying data augmentation boost your dataset to tenfold, preparing for training and real-time deployment.

Train a YOLOv4 object detector with the Darknet framework, preparing data and configuration files for CPU or GPU deployment, and test mask-detection on images, video, and webcam.

Configure the training files for a custom YOLO model in step one, adapting the Darknet config for two classes and creating a Drain and Distort metadata file.

Create the obj.names file to host class names for training, using the Sublime Text editor, saving as obj.names; the model will reference it during training before adding annotated image datasets.

Import edited and downloaded images with their label files into the darknet training dataset, creating a dataset folder and setting up metadata for train and test files.

Automatically generate train and test metafiles that list image paths, with an 80/20 train-test split, by running a Python script in the darknet data folder to prepare for training.

Train a YOLOv4 model by preparing the Darknet data folder with train and test images, configuring names and backup, and running Darknet to generate checkpoints and monitor GPU usage.

Experiment with training checkpoints to minimize loss, then test YOLOv4 on images and videos for mask detection using darknet commands, and save results to an output video.

Train on your own dataset by configuring sets one to five and the metafile, then test in real time on video or image to report detected classes and frame rate.

Set total iterations, such as 2000 per class, for efficient training. Monitor validation loss to prevent overfitting and use mean average precision to select the best weights for detection.

Review training of custom YOLOv4 models, adjust configuration files for number of glasses in the models, and edit or download datasets to train and test from scratch.

Explore building a graphical user interface for object detection with PyQt, using python bindings for the Cuty application development framework, install the toolkit, and add simple widgets for interactivity.

This lecture walks you through installing PyQt5 and Qt Designer tools, setting up the development environment, and launching Qt Designer to design user interfaces for your bike duty app.

Design a graphical user interface for an object detection app using a GUI designer, adding a main window with browse and run buttons, and a video path field.

Learn to implement real-time object detection with YOLOv4 using PyQt in the Cuty framework, including setup, file preparation, GUI integration, and running the app.

Explore how five Python files power a YOLOv4 object detection UI, detailing the UI dialogue window, file handling, start video, frame updates, and drawing bounding boxes.

Add gui widgets to display per-frame object counts from real-time detections, linking the interface to the detection code and showing counts on an lcd for cars, trucks, and people.

Add a horizontal slider to the YOLOv4 app to adjust the detection threshold in real time, map slider values to 0–1, and observe how detections respond.

Add a checkbox widget to filter detections by class in the YOLOv4 app, enabling on-the-fly filtering to show only people or cars, with adjustable detection sensitivity via a slider.

Add a real-time plot graph widget to visualize object counts over time in a YOLOv4 detection workflow, with custom Python logic updating live x and y data.

Integrate the social distancing script into the PyQt app, link widgets to show people at risk and results over time, and adjust the detection threshold for real-time monitoring.

Conclude the course by showing how to integrate python object detection code, deploy cross-platform apps, and train custom models for covid-19 projects and Facebook applications.

Build a real-time facial recognition attendance GUI in PyQt, featuring clock-in and clock-out with confirmation dialogs, live status, date and time displays, and a user-friendly interface designed with Qt Designer.