utils.evaluate.SODAeval

class SODAeval(labelmap: List[str])

Bases: object

Model evaluation tool

The add method accumulates predictions. The get_eval method can be used to obtain current results for the accumulated predictions.

The COCO API is used for evaluation.

Parameters:

labelmap (List[str]) – List of class names.

Methods

add	Add new predictions
get_eval	Calculates network scores and prints the results to the console
reset	Resets accumulated data

add(gts: Tensor, preds: Tensor, img: Tensor) → None

Add new predictions

Parameters:

• gts (torch.Tensor) –

  Ground Truth. Shape [batch, anchor, 5].

  One record contains(class id, xlu, ylu, xrd, yrd).

• preds (torch.Tensor) –

  Shape [batch, anchor, 6].

  One preds contains (class, iou, xlu, ylu, xrd, yrd)

• img (torch.Tensor) – Input img. Shape [batch, channels, h, w].

get_eval() → None

Calculates network scores and prints the results to the console

reset() → None

Resets accumulated data