Labelling image dataset for vehicle's body damage detection - Part 2

We want to label a second batch of 5000 images with a budget of approximately 500€. The dataset is oriented to detect external surface failures of vehicle's body (never inner). It contains a subset of 13 categories of possible defects on a car. Most of them happen on collision. Labelling instructions: -------------------------------------------------------------------------------------------- In this task, you will use a box tool to draw a tight rectangle around each defect on a vehicle in the images.  Make sure to use only one rectangle for each defect  Make sure to draw rectangles VERY tight in size respect to the defects  Mark the defect even if it is partially occluded  Take into account the priority list when there is an overlapping between them  If a vehicle is unclear or less than 10x10 pixels do not mark it  If there are no vehicles or there is no defect to box, you must click 'Nothing to box' Important!: When labelling, the following order of priority should be followed. The idea is to use different labels for defects that are in different zones. Defects in the same area should be labelled in this order of priority: - Break - Perforation - Wheel Damage - Tire Damage - Dent - Burn - Crack Glass - Missing Part - Stripe - Scratch - Impact Glass - Small impact - Extremely Dirty That is, if there is a "break", you do not need to label the "scratches" and "dents" associated with that "break." Or if there is a "dent" along with "stripe", simply label it as "dent" and the bounding box delimits the entire area affected by the defect. Required output: ---------------------------------------------------------------------------------------------------- The output should be presented in COCO format. The final output to submit should be a JSON file where all image ids are unique from 0-9999 (in case this task is done by different people all their labels should be reconditioned to fulfil this requirement). Category examples: ---------------------------------------------------------------------------------------------- To describe better the categories, please find attached a couple of images on each category.