danbenitezf / ObserveSign

Observe Sign

The purpose of this project is to ingest a tasks responses from the scale api and perform a quality checks on the data. The quality checks are based on the data schema and the data itself. The data then is saved as a txt file in the local machine.

Setup

Clone the repo and install the dependencies.

git clone https://github.com/danbenitezf/ObserveSign.git
cd ObserveSign

npm install

Reflections

With more time on this exercise I would have liked to have implemented the following:

• Unit testing
• More error handling
• More data validation
• More data quality checks

Regarding the quality checks, I would have liked to have implemented the following:

• Check if the annotation size is the same as the image size (width and height) and in the event it is return an error as the annotation is not valid.
• Check if the annotation is inside the image and in the event it is not return an error as the annotation is not valid.
• Check if the annotation is a valid polygon (geometry) and in the event it is not return an error as the annotation is not valid.
• Set a minimun and maximun threshold for the occlusion and truncation values as in those cases the annotation may not be valid or may not be useful for the model.
• Find a way to differentiate traffic lights from traffic signs as the annotation type is the same for both.
• Validate the background color for the traffic lights is set as "other"
• Analyze width and height from annotations to determine if the annotation is individual or should be splitted into multiple annotations.
• Refactor the code to make it more modular and reusable.
• check if the output is in the correct format to be used.