Add absolute error and MAE as custom metrics

[manushreegangwar]

This PR adds absolute_error and mean_absolute_error metric operators (similar to this)

import random
import numpy as np
import fiftyone as fo
import fiftyone.zoo as foz

dataset = foz.load_zoo_dataset("quickstart").select_fields().clone()

# Populate some fake regression data
for idx, sample in enumerate(dataset, 1):
    ytrue = random.random() * idx
    ypred = ytrue + np.random.randn() * np.sqrt(ytrue)
    confidence = random.random()
    sample["ground_truth"] = fo.Regression(value=ytrue)
    sample["predictions"] = fo.Regression(value=ypred, confidence=confidence)
    sample.save()

custom_metrics = {}
custom_metrics.update({"@voxel51/metric-examples/example_metric": dict(value="hello")})
custom_metrics.update({"@voxel51/metric-examples/absolute_error": {}})
custom_metrics.update({"@voxel51/metric-examples/mean_absolute_error": dict(error_eval_key="eval_absolute_error")})

results = dataset.evaluate_regressions(
    "predictions",
    gt_field="ground_truth",
    eval_key="eval",
    custom_metrics=custom_metrics,
    metric="absolute_error",
)

results.custom_metrics_report()
# {'Example metric': 'hello', 'Mean Absolute Error Metric': 5.0326606526479445}

This also works for video datasets.

dataset = foz.load_zoo_dataset("quickstart-video")

# Populate some fake regression data
for idx, sample in enumerate(dataset, 1):
    for fidx, frame in enumerate(sample.frames.values()):
        ytrue = random.random() * fidx
        ypred = ytrue + np.random.randn() * np.sqrt(ytrue)
        confidence = random.random()
        frame["ground_truth"] = fo.Regression(value=ytrue)
        frame["predictions"] = fo.Regression(value=ypred, confidence=confidence)
    sample.save()

results = dataset.evaluate_regressions(
    "frames.predictions",
    gt_field="frames.ground_truth",
    eval_key=eval_key,
    custom_metrics=custom_metrics,
    metric="absolute_error",
)