evaluator fix

[Zanna-93]

Description

1. Inside the Evaluator class constructor a KeyError Exception was raised when instantiating gt array because keys inside ground_thruth dictionary were absolute paths of images and the access was attempted using only the filename retrieved with key.split("/")[-1] (evaluator.py:59)
2. Restored previous allocation of merged_data dictionary because anAttributeError exception was raised when instruction gts = [list(item) for item in gt.xyxy] was executed, this because the gt object does not have an attribute named xyxy. That attribute exists in the predictions object

Comments left for in place explanations.

Type of change

Please delete options that are not relevant.

• [X ] Bug fix (non-breaking change which fixes an issue)

How has this change been tested, please provide a testcase or example of how you tested the change?

Execution of roboflow_example.py script against a YOLOv5 PyTorch (txt) dataset.

Docs

• Docs updated? Description of model_type CLI parameter and link to example

[CLAassistant]

All committers have signed the CLA.