Detection Workflow

class biapy.engine.detection.Detection_Workflow(cfg, job_identifier, device, args, **kwargs)[source]

Bases: Base_Workflow

Detection workflow where the goal is to localize objects in the input image, not requiring a pixel-level class. More details in our documentation.

Parameters:

cfg (YACS configuration) – Running configuration.
Job_identifier (str) – Complete name of the running job.
device (Torch device) – Device used.
args (argpase class) – Arguments used in BiaPy’s call.

define_metrics()[source]: Definition of self.metrics, self.metric_names and self.loss variables.

metric_calculation(output, targets, metric_logger=None)[source]

Execution of the metrics defined in define_metrics() function.

Parameters:

output (Torch Tensor) – Prediction of the model.
targets (Torch Tensor) – Ground truth to compare the prediction with.
metric_logger (MetricLogger, optional) – Class to be updated with the new metric(s) value(s) calculated.

Returns:

value – Value of the metric for the given prediction.

Return type:

float

detection_process(pred, filenames, metric_names=[], patch_pos=None, verbose=False)[source]

Detection workflow engine for test/inference. Process model’s prediction to prepare detection output and calculate metrics.

Parameters:

pred (4D Torch tensor) – Model predictions. E.g. (z, y, x, channels) for both 2D and 3D.
filenames (List of str) – Predicted image’s filenames.
metric_names (List of str) – Metrics names.

normalize_stats(image_counter)[source]

Normalize statistics.

Parameters:: image_counter (int) – Number of images to average the metrics.

after_merge_patches(pred)[source]

Steps need to be done after merging all predicted patches into the original image.

Parameters:: pred (Torch Tensor) – Model prediction.

process_patch(z, y, x, _filename, total_patches, c, pred, d, file_ext, z_dim, y_dim, x_dim)[source]

Process a patch for the detection workflow.

Parameters:

z (int) – Patch z index.
y (int) – Patch y index.
x (int) – Patch x index.
_filename (str) – Filename of the predicted image H5/Zarr.
total_patches (int) – Total number of patches.
c (int) – Current patch number.
pred (4D numpy array) – Model prediction.
d (int) – Number of digits of the total patches.
file_ext (str) – File extension of the predicted image.
z_dim (int) – Dimension of the z axis.
y_dim (int) – Dimension of the y axis.
x_dim (int) – Dimension of the x axis.

Returns:

df_patch (DataFrame) – Detected points in the patch.
fname (str) – Filename of the patch.

after_merge_patches_by_chunks_proccess_patch(filename)[source]

Place any code that needs to be done after merging all predicted patches into the original image but in the process made chunk by chunk. This function will operate patch by patch defined by DATA.PATCH_SIZE + DATA.PADDING.

Parameters:: filename (List of str) – Filename of the predicted image H5/Zarr.

process_sample(norm)[source]

Function to process a sample in the inference phase.

Parameters:: norm (List of dicts) – Normalization used during training. Required to denormalize the predictions of the model.

torchvision_model_call(in_img, is_train=False)[source]

Call a regular Pytorch model.

Parameters:

in_img (Tensor) – Input image to pass through the model.
is_train (bool, optional) – Whether if the call is during training or inference.

Returns:

prediction – Image prediction.

Return type:

Tensor

after_full_image(pred)[source]

Steps that must be executed after generating the prediction by supplying the entire image to the model.

Parameters:: pred (Torch Tensor) – Model prediction.

after_all_images()[source]: Steps that must be done after predicting all images.

print_stats(image_counter)[source]

Print statistics.

Parameters:: image_counter (int) – Number of images to call normalize_stats.

prepare_detection_data()[source]: Creates detection ground truth images to train the model based on the ground truth coordinates provided. They will be saved in a separate folder in the root path of the ground truth.