od-metrics 0.0.12

Object Detection metrics.

A metrics package for Object Detection.

Supported Metrics

Supported metrics include mAP (Mean Average Precision), mAR (Mean Average Recall) and Intersection over Union IoU.

Why OD-Metrics?

• User-friendly: simple to set and simple to use;
• Highly Customizable: every parameters that occur in the definition of mAP and mAR can be set by user to custom values;
• Compatibility with COCOAPI: each calculated metric is tested to coincide with COCOAPI metrics.

Installation

Install from PyPI

pip install od-metrics

Install from Github

pip install git+https://github.com/EMalagoli92/OD-Metrics

Documentation

For help, usage and API reference, please refer to Documentation

Simple Example

from od_metrics import ODMetrics

# Ground truths
y_true = [
    { # image 1
     "boxes": [[25, 16, 38, 56], [129, 123, 41, 62]],
     "labels": [0, 1]
     },
    { # image 2
     "boxes": [[123, 11, 43, 55], [38, 132, 59, 45]],
     "labels": [0, 0]
     }
    ]

# Predictions
y_pred = [
    { # image 1
     "boxes": [[25, 27, 37, 54], [119, 111, 40, 67], [124, 9, 49, 67]],
     "labels": [0, 1, 1],
     "scores": [.88, .70, .80]
     },
    { # image 2
     "boxes": [[64, 111, 64, 58], [26, 140, 60, 47], [19, 18, 43, 35]],
     "labels": [0, 1, 0],
     "scores": [.71, .54, .74]
     }
    ]

metrics = ODMetrics()
output = metrics.compute(y_true, y_pred)
print(output)
"""
{'mAP@[.5 | all | 100]': 0.2574257425742574,
 'mAP@[.5:.95 | all | 100]': 0.10297029702970294,
 'mAP@[.5:.95 | large | 100]': -1.0,
 'mAP@[.5:.95 | medium | 100]': 0.10297029702970294,
 'mAP@[.5:.95 | small | 100]': -1.0,
 'mAP@[.75 | all | 100]': 0.0,
 'mAR@[.5 | all | 100]': 0.25,
 'mAR@[.5:.95 | all | 100]': 0.1,
 'mAR@[.5:.95 | all | 10]': 0.1,
 'mAR@[.5:.95 | all | 1]': 0.1,
 'mAR@[.5:.95 | large | 100]': -1.0,
 'mAR@[.5:.95 | medium | 100]': 0.1,
 'mAR@[.5:.95 | small | 100]': -1.0,
 'mAR@[.75 | all | 100]': 0.0,
 'classes': [0, 1],
 'n_images': 2}
"""

Aknowledgment

• TorchMetrics
• COCO API
• Object-Detection-Metrics

License

This work is made available under the MIT License