Using Annotation GUI
Directory Structure for Data
Training data and annotations are organized as follows-
- training_data node is the ROS node which is supposed to contain all the training data.
- Folder training_data/preTrainData contains data collected for pre-training phase.
- This folder should have these 5 sub-folders-
- Annotations (can be a soft link to another drive e.g., /mnt/theakston/srvcData09/Annotations)
- Images (can be a soft link to another drive e.g., /mnt/theakston/srvcData09/data/Images)
- tmp
- DPMcache
- ImageSets (this folder stores files with list of training/test images)
Running Annotator
- Annotator can be launched using the command- $roslaunch DPM Annotation.launch
- User is prompted to enter a number between 0 and 9 to choose the category to annotate
- Once the GUI is visible. Go- File>open and choose the directory where data is stored
- Draw a bounding box by click and drag, press save and next(this actually saves the file)
- To Annotate multiple objects in the same image, press save after drawing a box and then draw another bounding box.
--
AnkurGupta - 12 Nov 2009
This topic: LCI
> WebHome >
TheSemanticRobotVisionChallenge > AnnotatingData
Topic revision: r1 - 2009-11-12 - AnkurGupta