Data for this competition consists of images along with bounding box annotations. The images are available in multiple sizes. Filenames take the form Image_ID.JPG. Train.csv contains the Image_IDs and bounding box annotations for the training set. SampleSubmission.csv shows the submission format and contains the Image_IDs of the test set.

Bounding boxes are rectangles defined by four numbers: x, y, w (width) and h (height). The rectangle spans from x to (x+w), and from y to (y+h). These numbers are floats (fractions of the image dimensions in pixels). The starter notebook shows how to convert these to pixel locations whatever the input image size is.

Wildbook and Wildbook-IA code are open source and free to use. Here is the link to Wildbook-IA on our github: https://github.com/WildbookOrg/wildbook-ia. You can use this as a starting point.