Welcome to the Computer Vision Project Tutorial! In this session, we'll talk about your computer vision projects. We will cover the creation of image datasets using annotation tools, a practical demonstration of one such tool, an overview of the YOLO (You Only Look Once) framework, and guidance on formatting and saving your dataset in YOLO's preferred format.
The assignment aims to apply the skills and knowledge you have acquired about image datasets, the YOLO object detection framework, and image annotation tools. You are required to either create a new image dataset or annotate an existing dataset, focusing on one of the following computer vision tasks: classification, detection, segmentation, or tracking.
- Choose a problem or topic in computer vision that interests you.
- The problem should align with one of the following tasks:
- Classification
- Detection
- Segmentation
- Pose estimation
- Tracking
-
Option 1 - Create New Data:
-
Collect and assemble a dataset of images relevant to your chosen problem.
-
Ensure the dataset is diverse and representative of the problem.
-
refer to canvas to get more details
-
-
Option 2 - Annotate Existing Data:
-
Select an existing vision dataset.
-
Create new labels or annotations for a subset of this dataset.
-
Refer to canvas for more details
-
-
Clearly define the labels or annotations you will use.
-
For detection, segmentation, or tracking tasks, annotate each image with bounding boxes or segmentation masks, as appropriate.
-
Ensure annotations are accurate and consistent.
Refer to canvas assignments to get better understanding of submission format.
You can use one of the following tools for annotation:
-
Label Studio
Description: Label Studio is an open source data labeling tool that supports multiple projects, users, and data types in one platform. It allows you to perform different types of labeling with many data formats. Also, you can integrate Label Studio with machine learning models to supply predictions for labels (pre-labels).
Website: Label Studio
Tutorials: Text: Label Studio Documentation — Overview of Label Studio
Video: Quickly Create Datasets for Training YOLO Object Detection with Label Studio | Label Studio
-
Makesense.ai
Description: Makesense.ai is a free, user-friendly online tool for labeling photos, ideal for quick annotation tasks.
Website: Makesense.ai
GitHub: SkalskiP/make-sense: Free to use online tool for labelling photos. https://makesense.ai github.com
Tutorials:
Text: Annotate Your Image — Using Online Annotation Tool! | by Sabina Pokhrel | Towards Data Science
Video: https://youtu.be/7nYYb6vwCjM
-
VGG Image Annotator (VIA)
Description: VGG Image Annotator is a simple and standalone manual annotation software for image, audio and video. VIA runs in a web browser and does not require any installation or setup.
Website: Visual Geometry Group - University of Oxford
Tutorials:
-
Roboflow
Description: Roboflow primarily serves as an annotation tool for computer vision models. It streamlines the annotation process, allowing users to label and annotate images efficiently for tasks like object detection. Additionally, it integrates with popular frameworks for model training and deployment.
For individuals with personal projects, school assignments, or research projects, Roboflow offers the Public plan to encourage exploration of computer vision. All projects on the Public plan are shared on Roboflow Universe. They offer to work upto 10000 source images for annotation.
What i liked about Roboflow was that you can directly upload a video file and it will create images from it itself.
Website: Roboflow => Select public plan for free!
Tutorials:
Text: Introduction - Roboflow Docs