Annotations

The annotation service, which is part of TAO Data Services, offers tools for users to easily manipulate groundtruth labels. tao dataset annotations supports the following tasks:

• convert

• slice

• merge

These tasks can be invoked from the TAO Toolkit Launcher using the following convention on the command-line:

            

            
tao dataset annotations <sub_task> <args_per_subtask>
        

Where args_per_subtask are the command-line arguments required for a given subtask. Each subtask is explained in the following sections.

Currently, the annotation service supports the KITTI and COCO formats.

The following is a sample spec file for converting a COCO dataset to KITTI format. It has three key components–data, kitti, and coco–as well as a global parameter, all of which are described below.

            

            
data:
  input_format: "COCO"
  output_format: "KITTI"
  output_dir: "/workspace/output"
kitti:
  image_dir: "/workspace/kitti/images"
  label_dir: "/workspace/kitti/labels"
  mapping: "/workspace/kitti_mapping.json"
coco:
  ann_file: "/workspace/coco.json"
results_dir: "/path/to/results"
        

Data Config

The data configuration (data) specifies the source and target formats of the label conversion, as well as the output path.

KITTI Config

The KITTI configuration (kitti) specifies the KITTI dataset information.

Here is an example of a category mapping file:

            

            
- person:
  - person
  - Person
  - person_group
  - rider
- bag:
  - hand_bag
  - backpack
  - personal_bag
- face:
  - face
        

COCO Config

The COCO configuration (coco) specifies the COCO annotation file location.

The annotation conversion service can be invoked from the TAO Toolkit Launcher using the following convention on the command-line:

            

            
tao dataset annotations convert [-h] -e <experiment spec>
                                [-r <results_dir>]
        

Required Arguments

• -e, --experiment_spec_file: The experiment specification file

Optional Arguments

• -h, --help: Show this help message and exit.

Here’s an example of using the convert command in Data Services:

Example

            

            
tao dataset annotations convert -e /path/to/spec.yaml
        

The following is a sample spec file for slicing a COCO annotation file. It has two key components–data and filter–as well as a global parameter (results_dir), all of which are described below.

            

            
data:
  annotation_file: /datasets/coco/annotations/instances_val2017.json
filter:
  mode: "category" # random, number
  num_samples: 10
  split: 5
  excluded_categories:
    - person
results_dir: /output/dir/
        

Data Config

The dataset configuration (data) specifies the input format and annotation file.

Filter Config

The filter configuration (filter) specifies how the annotation data will be sliced, which can be done in one of four modes:

1. random: Randomly split the annotation file into N partitions or sample the annotation file by a certain percentage

2. category: Filter annotation labels by the desired categories

3. number: Pick N samples in order from the annotations

4. filename: Filter the annotations by their file names

The annotation slicing service can be invoked from the TAO Toolkit Launcher using the following convention on the command-line:

            

            
tao dataset annotations slice [-h] -e <experiment spec>
                               [-r <results_dir>]
        

Required Arguments

• -e, --experiment_spec_file: The experiment specification file

Optional Arguments

• -h, --help: Show this help message and exit.

Example

Here’s an example of using the slice command in Data Services:

            

            
tao dataset annotations slice -e /path/to/spec.yaml
        

The following is a sample spec file for merging COCO annotation files. It has two key components–data and filter–as well as a global parameter (results_dir), all of which are described below.

            

            
data:
  format: "COCO"
  annotations:
    - /datasets/part_0.json
    - /datasets/part_1.json
    - /datasets/part_2.json
    - /datasets/part_3.json
    - /datasets/part_4.json
results_dir: /output/dir/
        

Data config

The data configuration (data) specifies the input format and annotatioin file.

The annotation merging service can be invoked from the TAO Toolkit Launcher using the following convention on the command-line:

            

            
tao dataset annotations merge [-h] -e <experiment spec>
                               [-r <results_dir>]
        

Required Arguments

• -e, --experiment_spec_file: The experiment specification file

Optional Arguments

• -h, --help: Show this help message and exit.

Example

Here’s an example of using the merge command in Data Services:

            

            
tao dataset annotations merge -e /path/to/spec.yaml