Applies image cutout augmentation https://arxiv.org/abs/1708.04552 Args: im (torch.Tensor): 3-D tensor to be augmented. labels (L

Apply element-wise weight and reduce loss. Args: loss (Tensor): Element-wise loss. weight (Tensor): Element-wise weights.

Convert nx6 boxes from [z, x, y, d, w, h] normalized to [z1, x1, y1, z2, x2, y2] where zxy1=top-left, zxy2=bottom-right Args: labels (tor

Convert nx6 boxes from [z1, x1, y1, z2, x2, y2] to [z, x, y, d, w, h] normalized where zxy1=top-left, zxy2=bottom-right Args: labels (tor

Calculate losses Args: p (torch.Tensor): predictions to test targets (torch.Tensor): targets corresponding to predict

Loads niftis and converts to torch tensor to be fed as input to the model. Args: index (int): dataset index of image to be read.

Initialization for YOLO loss. Args: model (torch.Module): YOLO model to calculate loss for. autobalance (bool, option

WIP - doesn't fill confusion matrices, could be due to bad predictions Create confusion matrices to save training results Updated

Initialization for the inference Dataset Args: path (str): parent directory for the Dataset's files img_size (int, op

Initialization for the training Dataset Args: path (str): parent directory for the Dataset's files img_size (int, opt

3D convolution and 3D batchnorm layer for 3D YOLO models Args: c1 (int): channels in c2 (int): channels out

Bottleneck layers for 3D YOLO models Args: c1 (int): channels in c2 (int): channels out shortcut (bool, o

Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher Args: c1 (int): channels in c2 (int): channels

YOLO Detection layer, adjusted for 3D Args: nc (int, optional): Number of classes. Defaults to 80. anchors (Tuple[int

3D YOLO model base class. Args: cfg (Dict[str] or str, optional): Model configuration dictionary or path to yaml file containing

Check anchor order against stride order for YOLO3D Detect() module m, and correct if necessary. Suggested not to run Reverses an

Generates the dataset dictionary used during training. Has removed download functionality found in YOLOv5 version. Args: data (str, p

Used to collate images to create the input batches

Returns all the registered actions by callback hook Args: hook The name of the hook to check, defaults to all

Register a new action to a callback hook Args: hook The callback hook name to register the action to

Scales img(bs,1,z,y,x) by ratio optionally constrained to a multiple of gs Args: img (torch.Tensor): image to be resized ratio (f

Converts unnormalized torch tensor into a nifti image. Useful for debugging new (e.g. non-NIfTI) dataloader pipelines.

Verify one image-label pair. Works for .nii and .nii.gz files. Args: args (Tuple[str]): contains the image path, label path, and error m