tensorflow/sagemakercv/data/coco/coco_metric.py [372:401]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        predictions, self._include_mask, is_image_mask=is_predict_image_mask)
    coco_eval = COCOeval(self.coco_gt, coco_dt, iouType='bbox')#, use_ext=True, num_threads=32) 
    coco_eval.params.imgIds = image_ids
    coco_eval.evaluate()
    coco_eval.accumulate()
    coco_eval.summarize()
    coco_metrics = coco_eval.stats

    if self._include_mask:
      # Create another object for instance segmentation metric evaluation.
      mcoco_eval = COCOeval(self.coco_gt, coco_dt, iouType='segm')#, use_ext=True, num_threads=32)
      mcoco_eval.params.imgIds = image_ids
      mcoco_eval.evaluate()
      mcoco_eval.accumulate()
      mcoco_eval.summarize()
      mask_coco_metrics = mcoco_eval.stats

    if self._include_mask:
      metrics = np.hstack((coco_metrics, mask_coco_metrics))
    else:
      metrics = coco_metrics

    # clean up after evaluation is done.
    self._reset()
    metrics = metrics.astype(np.float32)

    metrics_dict = {}
    for i, name in enumerate(self.metric_names):
      metrics_dict[name] = metrics[i]
    return metrics_dict
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



tensorflow/sagemakercv/data/coco/coco_metric.py [413:442]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
        predictions, self._include_mask, is_image_mask=is_predict_image_mask)
    coco_eval = COCOeval(self.coco_gt, coco_dt, iouType='bbox') #, use_ext=True, num_threads=32)
    coco_eval.params.imgIds = image_ids
    coco_eval.evaluate()
    coco_eval.accumulate()
    coco_eval.summarize()
    coco_metrics = coco_eval.stats

    if self._include_mask:
      # Create another object for instance segmentation metric evaluation.
      mcoco_eval = COCOeval(self.coco_gt, coco_dt, iouType='segm') # , use_ext=True, num_threads=32)
      mcoco_eval.params.imgIds = image_ids
      mcoco_eval.evaluate()
      mcoco_eval.accumulate()
      mcoco_eval.summarize()
      mask_coco_metrics = mcoco_eval.stats

    if self._include_mask:
      metrics = np.hstack((coco_metrics, mask_coco_metrics))
    else:
      metrics = coco_metrics

    # clean up after evaluation is done.
    self._reset()
    metrics = metrics.astype(np.float32)

    metrics_dict = {}
    for i, name in enumerate(self.metric_names):
      metrics_dict[name] = metrics[i]
    return metrics_dict
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -