Multiview images

Introduction

This easy-to-follow tutorial will show you how to upload multiview images and label groups to Supervisely using Python SDK and get the advantage of the multiview image annotaion in the Supervisely Labeling Toolbox, which allows you to label images quickly and efficiently on one screen. You will learn how to enable multiview in the project settings, upload multiview images and explore the multiview in the labeling interface.

How to debug this tutorial

Step 1. Prepare ~/supervisely.env file with credentials. Learn more here.

Step 2. Clone the repository with source code and demo data and create a Virtual Environment.

git clone https://github.com/supervisely-ecosystem/import-multiview-images-tutorial.git

cd import-multiview-images-tutorial

sh create_venv.sh

Step 3. Open the repository directory in Visual Studio Code.

code -r .

Step 4. Change the Workspace ID in the local.env file by copying the ID from the context menu.

WORKSPACE_ID=942 # ⬅️ change value

Step 5. Start debugging src/main.py.

Import libraries

import os

from dotenv import load_dotenv

import supervisely as sly

Load environment variables

if sly.is_development():
    load_dotenv("local.env")
    load_dotenv(os.path.expanduser("~/supervisely.env"))

workspace_id = sly.env.workspace_id()

Init API client

api = sly.Api.from_env()

Explore the directory with images

here is the structure of the directory with images (src/images):

 📂 images
 ┣ 📂 audi
 ┃ ┣ 🏞️ audi_01.jpg
 ┃ ┣ 🏞️ audi_02.jpg
 ┃ ┗ 🏞️ audi_03.jpg
 ┣ 📂 mercedes
 ┃ ┣ 🏞️ mercedes_01.jpg
 ┃ ┣ 🏞️ mercedes_02.jpg
 ┃ ┣ 🏞️ mercedes_03.jpg
 ┃ ┣ 🏞️ mercedes_04.jpg
 ┃ ┣ 🏞️ mercedes_05.jpg
 ┃ ┗ 🏞️ mercedes_06.jpg
 ┣ 📂 renault
 ┃ ┣ 🏞️ renault_01.jpg
 ┃ ┣ 🏞️ renault_02.jpg
 ┃ ┣ 🏞️ renault_03.jpg
 ┃ ┗ 🏞️ renault_04.jpg
 ┗ 📂 ford
   ┣ 🏞️ ford_01.jpg
   ┣ 🏞️ ford_02.jpg
   ┣ 🏞️ ford_03.jpg
   ┣ 🏞️ ford_04.jpg
   ┗ 🏞️ ford_05.jpg

Create a new project and dataset

project = api.project.create(workspace_id, "Grouped cars", change_name_if_conflict=True)
dataset = api.dataset.create(project.id, "ds0")

Enable multiview in the project settings

api.project.set_multiview_settings(project.id)

You can also enable multiview in the Image Labeling Tool interface:

And now we're ready to upload images.

How to upload multiview images

In this tutorial, we'll be using the api.image.upload_multiview_images method to upload multiview images to Supervisely.

def upload_multiview_images(
    dataset_id: int,
    group_name: str,
    paths: Optional[List[str]] = None,
    metas: Optional[List[Dict]] = None,
    progress_cb: Optional[Union[tqdm, Callable]] = None,
    links: Optional[List[str]] = None,
    conflict_resolution: Optional[Literal["rename", "skip", "replace"]] = "rename",
    force_metadata_for_links: Optional[bool] = False,
) -> List[ImageInfo]:

So, the method uploads images to Supervisely and returns a list of ImageInfo objects.

Upload multiview images

for group_dir in os.scandir("src/images"):
    if not group_dir.is_dir():
        continue
    images_paths = sly.fs.list_files(group_dir.path, valid_extensions=sly.image.SUPPORTED_IMG_EXTS)

    api.image.upload_multiview_images(dataset.id, group_dir.name, images_paths)

Group existing images for multiview

If you already have images uploaded to Supervisely and you want to group them for multiview, you can use the api.image.group_images_for_multiview method.

images = [2389126, 2389127, 2389128, 2389129, 2389130, ...]

for idx, batch_ids in enumerate(sly.batched(images, batch_size=6)):
    api.image.group_images_for_multiview(batch_ids, f"group_{idx}")

Grouped view in the labeling interface

So now, that we've uploaded all the images, let's take a look at the labeling interface.

As you can see, the images in the Labeling tool are grouped in the same way as in your images in folders (images from one folder are combined into one group). When importing, each image from the folders will be assigned tags with the same values, which allows them to be grouped into one group.

Multiview labeling can be very useful when annotating objects of multiple classes simultaneously on several images. You don't need to shift your attention to find the necessary class every time you switch between images, allowing you to increase efficiency and save time and effort.

How to upload label groups

There are many cases when you need to group labels together. For example, if you have some labels captured from different perspectives that represent one object on different images and you want to analyze the object as a whole and not as separate instances, you can join them into a single group.

Label group - is a simple group of objects, that displays the relationship between objects and helps you to quickly locate the object on different images and to avoid labeling the same object multiple times.

Using the api.annotation.append_labels_group method, you can upload labels as a group to images.

def append_labels_group(
    self,
    dataset_id: int,
    image_ids: List[int],
    labels: List[Label],
    project_meta: Optional[ProjectMeta] = None,
    group_name: Optional[str] = None,
) -> None:

Let's group it all together and upload local images and labels to Supervisely using this method.

Our sample data directory structure:

 📂 data
 ┣ 📂 images
 ┃ ┣ 🏞️ car_01.jpeg
 ┃ ┣ 🏞️ car_02.jpeg
 ┃ ┗ 🏞️ car_03.jpeg
 ┗ 📂 masks
   ┣ 🏞️ car_01.png
   ┣ 🏞️ car_02.png
   ┗ 🏞️ car_03.png

⬇️ You can download this sample here: data.zip

Follow the code below to upload images and labels to Supervisely.

project_id = 56
dataset_id = 196

# GET PROJECT META
meta = sly.ProjectMeta.from_json(api.project.get_meta(project_id, with_settings=True))

# GET OBJ CLASS FROM META BY NAME
obj_cls = meta.get_obj_class("car")
# OR CREATE NEW OBJ CLASS IF NOT EXISTS
# obj_cls = sly.ObjClass(name="car", geometry_type=sly.Rectangle, color=[255, 0, 0])
# UPDATE PROJECT META IF CREATING NEW OBJ CLASS
# meta = meta.add_obj_classes([obj_cls])
# api.project.update_meta(project_id, meta.to_json())

# SET MULTIVIEW SETTINGS
api.project.set_multiview_settings(project_id)

# GET IMAGES AND MASKS PATHS
image_dir = os.path.join("data", "images")
mask_dir = os.path.join("data", "masks")

# SORT PATHS FOR CORRECT LABELS ORDER
image_paths = sorted([os.path.join(image_dir, path) for path in os.listdir(image_dir)])
mask_paths =  sorted([os.path.join(mask_dir, path) for path in os.listdir(mask_dir)])

# CREATE LABELS
labels = []
for image_path, mask_path in zip(image_paths, mask_paths):
    # READ MASK
    bitmap = sly.Bitmap.from_path(mask_path)
    # CREATE LABEL
    label = sly.Label(geometry=bitmap, obj_class=obj_cls)
    labels.append(label)

# UPLOAD IMAGES
image_infos = api.image.upload_multiview_images(dataset_id, "white_car", image_paths)
images_ids = [image_info.id for image_info in image_infos]

# APPEND LABELS TO IMAGES
api.annotation.append_labels_group(
    dataset_id=dataset_id,
    image_ids=images_ids,
    labels=labels,
    project_meta=meta,
)

Summary

In this tutorial, you learned how to upload multiview images and label groups to Supervisely using Python SDK and get the advantage of the multiview image annotation in the labeling interface, which allows you to label images quickly and efficiently on one screen. Let's recap the steps we did:

1. Create a new project and dataset.

2. Set multiview settings for the project using the api.project.set_multiview_settings method.

3. Upload images using the api.image.upload_multiview_images method.

4. Group existing images for multiview using the api.image.group_images_for_multiview method.

5. Upload label groups using the api.annotation.append_labels_group method.

And that's it! Now you can upload your multview images to Supervisely using Python SDK.