Description
The 2D bounding box tool is a type of annotation tool used in computer vision and machine learning applications to label and annotate objects of interest in a 2D image. A bounding box is a rectangle that encloses an object of interest, typically defined by its four corner points or by its center point and its width and height. The 2D bounding box tool allows users to draw bounding boxes around objects in an image and label them according to their class (e.g. car, pedestrian, traffic sign, etc.).
The 2D bounding box tool is commonly used in applications such as object detection and tracking, where the goal is to identify and locate objects of interest in an image or video stream. By labeling objects with bounding boxes, machine learning algorithms can be trained to recognize and classify objects in real-world scenarios.
In the context of the HyperSpec Road Mentor, the 2D bounding box tool can be used to label and annotate various objects in road scenes, such as vehicles, pedestrians, traffic signs, and road markings. These annotations can then be used to train and improve machine-learning models.
How to access 2D bounding box Tool:
User Interface
The bounding box tool interface consists of several components, including:
Image display: This displays the image or video frame to be annotated. in three different views left, center, and right view.
Bounding box selection: This allows the user to draw a bounding box around the object of interest using a cursor. this selection tool is available in all three different camera views.
Labeling options: This enables the user to assign a label or a category to the object within the bounding box. This is useful in object detection and classification tasks.
Navigation controls: These allow the user to move between frames or images to annotate them sequentially. On the top of the camera view, there is a playback bar.
Editing options: This enables the user to modify or delete existing bounding boxes, change labels, or adjust the size and position of the boxes.
Save options: This allows the user to save the annotated data.
Overall, the bounding box tool interface is designed to be user-friendly and intuitive, allowing even novice users to accurately and efficiently annotate objects in images and videos.
Process:
- Open the bounding box tool using the road mentor dashboard. Users can go to the trip database and open any trip in the tool.
- There will be a preloaded image, as the user will select the trip first.
- Once the images are loaded into the tool, the user can see pre-created bounding boxes in the dataset.
- If there are if a bounding box is missing for an object in an image or video, the user can create a new bounding box around the object using the bounding box tool.
- This is an important step in the labeling process to ensure that all objects in the image or video are properly annotated and labeled.
- Draw the bounding box: Use your mouse to draw a box around the object that you want to label. The box should be as tight as possible around the object without including any background pixels. There are multiple drawing tools are available in every camera image, these tools can be used as per requirement.
- Label the object: Assign a label to the object that you have selected, such as "car", "person", or "dog". You may also assign additional attributes to the object, such as its color or size.
- Save the annotation: Save the image with the bounding box annotation, along with any additional metadata that you have assigned to the object.
- Repeat: Repeat the process for each object in the image that you want to label. It is important to be consistent and accurate when using a bounding box tool. The accuracy of the annotations will impact the performance of machine learning models trained on the data.
- Users can use the previous and next buttons to navigate between the frames.
- The tool will propagate previous annotations and labels for the same object.
- If user feels there are a few wrong annotations, they can edit/correct as per the evidence as well.