Compute Detection Metrics: Precision and Recall

[jacobson15p]

Does the detection metrics library support computing precision and recall?

[hfslyc]

Hi,

We do support returning all precision points in PR curve in the metrics library. Please see https://github.com/waymo-research/waymo-open-dataset/blob/master/waymo_open_dataset/metrics/ops/detection_metrics_ops_test.py#L153 for an example.

[JennySeidenschwarz]

Is there a possibility to get PR and RE also from waymo_open_dataset/metrics/tools/compute_detection_metrics_main.cc ? Tried adding it to the output string << breakdown_names[i] << ": [mAP " << metric.mean_average_precision() << "]" << " [mAPH " << metric.mean_average_precision_ha_weighted() << "]" << " [PR " << metric.precisions() << "]" << " [RE " << metric.recalls() << "]\n";

but apparently I can not build it due to the following error: external/com_google_protobuf/src/google/protobuf/stubs/stringpiece.h:432:22: note: no known conversion for argument 2 from 'const google::protobuf::RepeatedField<float>' to 'google::protobuf::StringPiece'