vodet.detect module¶

class vodet.detect.Detector(classifier, classes, labels_path, label_type='labelme', conf_th=0.95, iou_th=0.3, step_ratio=0.5, input_size=[24, 24])¶

Bases: object

The detector object.

classifier¶

The GMVAE’s classifier.

Type: vodet.distributions.Classifier

classes¶

A dict of names and index of each classes. The keys must be the name of classes. The values must be the index of each classes (used by the classifier).

Type: dict of int

labels_path¶

The path for the annotation data of train dataset. If label_type is “VoTT”, this must be a path for the exported csv file. If label_type is “labelme”, this must be a path for the directory that contains labelme’s json output files.

Type: str

label_type¶

The type of label data, either “VoTT” (for VoTT’s csv export) or “labelme” (for labelme’s json export).

Type: str

conf_th¶

The confidence threshold for each proposed bounding box.

Type: float default 0.99

iou_th¶

The threshold of IoU value for Non-Maximum Supression of bounding boxes.

Type: float default 0.3

step_ratio¶

The ratio between the step size and width or height of sliding windows.

Type: float default 0.5

input_size¶

The input size of the classifier

Type: list of int

detect_dir(in_dir, out_dir)¶

Detect objects and save the results from images in a directory.

Parameters

in_dir (str) – The path of input directory.
out_dir (str) – The path of output directory.

Returns

df – A dataframe of detected objects.

Return type

pd.DataFrame

detect_img(in_path, out_path)¶

Detect objects and save the result from an image.

Parameters

in_path (str) – The path of input image.
out_path (str) – The path of output image.

Returns

nums – The detected numbers of each classes.

Return type

dict

draw_barplot(date_df, out_path)¶

Draw a bar plot of detected result. The x-axis stands for the date of each picture was taken and the y-axis stands for the detected numbers of each class. Run this after detect_dir.

Parameters

date_df (pd.DataFrame) – A dataframe that contains column “image” for the names of each image file, column “date” for the date each picture was taken. You can make this with vodet.utils.exif_date.
out_path (str) – The output path for the barplot.

Returns

df – A dataframe of detected result with “date” column.

Return type

pd.DataFrame

vodet.detect.detect(classifier, image_path, labels_path, out_path, classes, label_type, bb_color, conf_th=0.99, iou_th=0.3, step_ratio=0.5, input_size=[24, 24])¶

Detect object with GMVAE’s classifier and sliding windows.

Parameters

classifier (vodet.distributions.Classifier) – A trained Classifier object.
image_path (str) – The path for the target image.
labels_path (str) – The path for the annotation data of train dataset. If label_type is “VoTT”, this must be a path for the exported csv file. If label_type is “labelme”, this must be a path for the directory that contains labelme’s json output files.
out_path (str) – The path for the output image (input image with detected results).
classes (dict of int) – A dict of names and index of each classes. The keys must be the name of classes. The values must be the index of each classes (used by the classifier).
label_type (str) – The type of label data, either “VoTT” (for VoTT’s csv export) or “labelme” (for labelme’s json export).
bb_color (dict of tuple) – The RGB bounding box color values for each classes. RGB values should be 0-255 int.
conf_th (float default 0.99) – The confidence threshold for each proposed bounding box.
iou_th (float default 0.3) – The threshold of IoU value for Non-Maximum Supression of bounding boxes.
step_ratio (float default 0.5) – The ratio between the step size and width or height of sliding windows.
input_size (list of int) – The input size of the classifier

Returns

detected_numbers – The detected numbers of each classes

Return type

dict of int

vodet.detect.detect_old(classifier, image_path, labels_path, out_path, classes, label_type, bb_color, conf_th=0.99, iou_th=0.3, step_ratio=0.5, input_size=[24, 24], device='cuda:0')¶

Detect object with GMVAE’s classifier and sliding windows.

Parameters

classifier (vodet.distributions.Classifier) – A trained Classifier object.
image_path (str) – The path for the target image.
labels_path (str) – The path for the annotation data of train dataset. If label_type is “VoTT”, this must be a path for the exported csv file. If label_type is “labelme”, this must be a path for the directory that contains labelme’s json output files.
out_path (str) – The path for the output image (input image with detected results).
classes (dict of int) – A dict of names and index of each classes. The keys must be the name of classes. The values must be the index of each classes (used by the classifier).
label_type (str) – The type of label data, either “VoTT” (for VoTT’s csv export) or “labelme” (for labelme’s json export).
bb_color (dict of tuple) – The RGB bounding box color values for each classes. RGB values should be 0-255 int.
conf_th (float default 0.99) – The confidence threshold for each proposed bounding box.
iou_th (float default 0.3) – The threshold of IoU value for Non-Maximum Supression of bounding boxes.
step_ratio (float default 0.5) – The ratio between the step size and width or height of sliding windows.
input_size (list of int) – The input size of the classifier
device (str default "cuda:0") – The device name to use for running the classifier.

Returns

detected_numbers – The detected numbers of each classes

Return type

dict of int

vodet.detect.iou(a, b)¶

Intersection over Union

Parameters

a (tuple of int) – A tuple of size 4. Each element stands for x1, y1, x2, y2 of the bounding box
b (tuple of int) – A tuple of size 4. Each element stands for x1, y1, x2, y2 of the bounding box

Returns

iou – The calculated IoU value.

Return type

float

vodet.detect.nms(bboxes: list, scores: list, labels: list, iou_threshold: float) -> (<class 'list'>, <class 'list'>)¶

Non-Maximum Suppression

Parameters

bboxes (list of tuple) – A list of bounding boxes each contains x1, y1, x2, y2.
scores (list of float) – A list of confidence scores of each bounding box.
labels (list of str) – A list of label names of each bounding box.

Returns

new_bboxes (list of tuple) – A list of survived bounding boxes.
new_labels (list of str) – A list of label names for new_bboxes.