semlog_vis API

semlog_vis.image

image.cut_object(rgb_image_path, mask_image_path, object_color, saving_path=None, flag_remove_background=False, new_background_color=(255, 255, 255))

This function is used to cut a specific object from the pair RGB/mask image.

Parameters

• rgb_image_path – Path to a RGB image.

• mask_image_path – Path to a mask image.

• object_color – Mask color for the target object

• saving_path – Path to save the cropped image.

• flag_remove_background – A boolean whether to remvoe the background of the cropped image.

• new_background_color – If remove the background, use this color to fill the background.

Returns

Coordinates of the object in the image.

image.load_img_and_mask(img, mask)

Load a pair of image and its mask, convert BGR to RGB of the mask image.

image.pad_image(img_path, width, height, pad_type, value=(0, 0, 0))

Pad the give image with different size and padding type.

Parameters

• img_path – Path of the source image.

• width – The target width.

• height – The target height.

• pad_type – The padding type in cv2.copyMakeBorder.

• value – If pad_type=”cv2.Border_CONSTANT”, this value is used as the padding constant color.

Returns

The processed image.

image.remove_image_background(img, mask, mask_color, background_color=(255, 255, 255))

Replace the image background with a single color. :param img: The source image. :param mask: The source mask. :param mask_color: Mask color of the object. :param background_color: New background color to be replaced.

Returns

The new image as ndarray.s

image.replace_single_color(img, color, new_color)

Replace one color in an image to another color.

Parameters

• img – The source image.

• color – The color in the image that you want to replace.

• new_color – New color to replace.

Returns

New image as ndarray.

image.resize_all_images(image_dir, width, height, resize_type)

Multiprocessing function for resize images.

Parameters

• image_dir – A dict of images to be resized.

• width – Target width.

• height – Target height.

• resize_type – Stretch or sclae depending on the input.

image.resize_image(img_path, width='', height='', type='cut')

Resize one image.

Parameters

• img_path – Path of the image

• width – Target width.

• height – Target height.

• type – ‘cut’ -> Stretch the image absolutely, otherwise scale the image by width or height.

image.scale_image(img_path, ratio)

Scale an image with a given ratio.

Parameters

• img_path – Path to the image.

• ratio – ratio to scale the image.

semlog_vis.point_cloud

class point_cloud.PointCloudGenerator(rgb_file, depth_file, focal_length, scalingfactor)

Class for generate point cloud from RGB and depth image pair.

calculate(flag_depth_conversion=False)

Calculate the 3D position according to the depth data.

depthConversion(PointDepth)

Adjust the relative depth in UE4.

save_npy(path, alpha=False)

Save the .npy file of the point cloud.

show_point_cloud()

Show the point cloud directly with open3d.

write_ply(path)

Save the point cloud as a .ply file.

class point_cloud.PointCloudGenerator(rgb_file, depth_file, focal_length, scalingfactor)

Class for generate point cloud from RGB and depth image pair.

calculate(flag_depth_conversion=False)

Calculate the 3D position according to the depth data.

depthConversion(PointDepth)

Adjust the relative depth in UE4.

save_npy(path, alpha=False)

Save the .npy file of the point cloud.

show_point_cloud()

Show the point cloud directly with open3d.

write_ply(path)

Save the point cloud as a .ply file.

semlog_vis.create_annotation

create_annotation.clean_df_for_annotation(df)

Remove unused columns and create number mapping to classes.

Parameters

df – Result df from calculate bounding box.

Returns

Cleaned df. class_mapping_dict: Map class name to number.

Return type

df

create_annotation.convert_df_to_annotation(df, bb_path, mapping_path)

Convert a result df to annotation and store data as txt file

Parameters

• df – Result df from calcualte bounding box.

• bb_path – path to save bounding box data.

• mapping_path – path to save class mapping dict.

create_annotation.write_bb_to_txt(result, txt_path)

Wrtie bounding boxes to txt file

Parameters

• result – A result df contains bb information.

• txt_path – Path to save the data.

create_annotation.write_mapping_to_txt(class_mapping_dict, mapping_path)

Write mapping dict to txt.

Parameters

• class_mapping_dict – A dict maps class to number.

• mapping_path – Path to save the txt file.

Returns:

semlog_vis.bounding_box

bounding_box.calculate_bounding_box(df, object_rgb_dict, root_folder_path, root_folder_name)

Main function for calculate bounding boxes.

Parameters

• df – Information Data Frame.

• object_rgb_dict – A dict maps object id to mask colors.

• root_folder_path – Root path for images.

• root_folder_name – Root folder name.

bounding_box.crop_with_all_bounding_box(object_rgb_dict, image_dir)

Crop full images with max boundary of all objects.

Parameters

• object_rgb_dict – A dict contains objects and their mask colors.

• image_dir – A dict contains all target images

bounding_box.download_bounding_box(df, object_rgb_dict, root_folder_path, root_folder_name)

Main function for download bounding boxes.

Parameters

• df – Information Data Frame.

• object_rgb_dict – A dict maps object id to mask colors.

• root_folder_path – Root path for images.

• root_folder_name – Root folder name.