asyncbridge/pycococreator

Helper functions to create COCO datasets

Show bounding boxes and labels

I added codes for showing the bounding boxes and labels. To show it, I used Jupyter Notebook and created deployer.ipynb and visualize_cg.ipynb files.

• Directory Structure

    |__ examples					
        |__ crestedgecko
        	|__ deployer.ipynb
        	|__ visualize_cg.ipynb
            	|__ train
            		|__ annotations
            			|__ CG_train2018_000000000001_patternless_0.png
            			|__ instances_crestedgecko_train2018
            		|__ crestedgecko_train2018
            			|__ CG_train2018_000000000001.jpg
        |__ shapes
        	|__ ...	        

• deployer.ipynb

1. Rename and resize image files.
2. Create an annotation file from images and masking images.

• visualize_cg.ipynb

Show bounding boxes and labels including segmentations from the created annotation file.

Install

1. pip install git+git://github.com/waspinator/pycococreator.git@0.2.0
2. Run Jupyter Notebook.(Set working directory as pycococreator)
3. Go to pycococreator/examples/crestedgecko directory.
4. Open deployer.ipynb and run it.

• instances_crestedgecko_train2018.json annotation file will be created in pycococreator/examples/crestedgecko/train directory.

5. Copy the annotation file to pycococreator/examples/crestedgecko/annotations directory.
6. Open visualize_cg.ipynb and run it to show the sample image.

• Segmentation, bounding box and label will be drawn on the sample image.

License

This project is made available under the Apache 2.0 License.

It is forked from: https://github.com/waspinator/pycococreator

Reference

[1] https://patrickwasp.com/create-your-own-coco-style-dataset