torchmetrics-ext 0.1.1

An extension of torchmetrics package.

TorchMetrics Extension

Installation

Simple installation from PyPI

pip install torchmetrics-ext

What is TorchMetrics Extension

It is an extension of torchmetrics containing more metrics for machine learning tasks. Currently, it offers metrics for:

• 3D Visual Grounding
  • ScanRefer

Using TorchMetrics Extension

Here are examples for using the metrics in TorchMetrics Extension:

ScanRefer

It measures the thresholded accuracy Acc@kIoU, where the positive predictions have higher intersection over union (IoU) with the ground truths than the thresholds. The metric is based on the ScanRefer.

import torch
from torchmetrics_ext.visual_grounding import ScanReferMetric
metric = ScanReferMetric()
pred_aabbs = torch.tensor([[[0., 0., 0.], [1., 1., 1.]], [[0., 0., 0.], [2., 2., 2.]]], dtype=torch.float32)
gt_aabbs = torch.tensor([[[0., 0., 0.], [1., 1., 1.]], [[0., 0., 0.], [1.5, 1.5, 1.5]]], dtype=torch.float32)
gt_eval_types = ("unique", "multiple")
results = metric(pred_aabbs, gt_aabbs, gt_eval_types)