Make sure to change the trainingImagesDir path to the path for your own training images and the boxesFile to the path of the CSV file containing the necessary bounding boxes, otherwise the code will fail. The code also only uses bounding boxes with the label called car(line 29), so make sure to change the value to the type of object you are trying to detect.
This repo was made to convert a CSV file into an Edge Impulse bounding_boxes.labels (JSON) file.
The CSV was made with the columns that appeared like so:
| image | xmin | ymin | xmax | ymax |
|---|---|---|---|---|
| im001 | 281.259 | 187.0351 | 327.7279 | 223.2255 |
However, the code can be modified to fit whatever format you are dealing with,
as long as you are trying to create a bounding_boxes.labels file.