[Datumaro] Add merge command with segment intersection

[zhiltsov-max]

Motivation and context

• Added a command to merge any number of annotation datasets.
  • Annotations are matched by area, where it makes sense
  • Attributes and labels are merged by voting

If you want to test:

datum merge <project1> <project2> ...

Labels in the datasets must match.

Example (requires openvino):

interp.py

from datumaro.components.extractor import *

conf_thresh = 0.5

def process_outputs(inputs, outputs):
     # inputs = model input, array or images, shape = (N, C, H, W)
     # outputs = model output, shape = (N, 1, K, 7)
     # results = conversion result, [ [ Annotation, ... ], ... ]
     results = []
     for input, output in zip(inputs, outputs):
          input_height, input_width = input.shape[:2]
          detections = output[0]
          image_results = []
          for i, det in enumerate(detections):
               label = int(det[1])
               conf = float(det[2])
               if conf <= conf_thresh:
                    continue

               x = max(int(det[3] * input_width), 0)
               y = max(int(det[4] * input_height), 0)
               w = min(int(det[5] * input_width - x), input_width)
               h = min(int(det[6] * input_height - y), input_height)
               image_results.append(Bbox(x, y, w, h,
                    label=label, attributes={'score': conf} ))

               results.append(image_results)

     return results

def get_categories():
     label_categories = LabelCategories()
     label_categories.add('background')
     label_categories.add('face')
     return { AnnotationType.label: label_categories }

Face detection example

# create a project from some images with faces
datum project import -i images/ -f image_dir -o test_project
cd test_project

# annotate with models
# https://github.com/opencv/open_model_zoo/blob/master/models/intel/face-detection-adas-0001/model.yml
# https://github.com/opencv/open_model_zoo/blob/master/models/intel/face-detection-retail-0004/model.yml
# interp.py almost exactly like in the documentation https://github.com/opencv/cvat/blob/develop/datumaro/docs/user_manual.md#register-model
wget https://download.01.org/opencv/2020/openvinotoolkit/2020.3/open_model_zoo/models_bin/1/face-detection-adas-0001/FP32/face-detection-adas-0001.xml
wget https://download.01.org/opencv/2020/openvinotoolkit/2020.3/open_model_zoo/models_bin/1/face-detection-adas-0001/FP32/face-detection-adas-0001.bin
wget https://download.01.org/opencv/2020/openvinotoolkit/2020.3/open_model_zoo/models_bin/1/face-detection-retail-0004/FP32/face-detection-retail-0004.xml
wget https://download.01.org/opencv/2020/openvinotoolkit/2020.3/open_model_zoo/models_bin/1/face-detection-retail-0004/FP32/face-detection-retail-0004.bin
datum model add -l open_vino -- -w face-detection-retail-0004.bin -d face-detection-retail-0004.xml -i interp.py
datum model add -l open_vino -- -w face-detection-adas-0001.bin -d face-detection-adas-0001.xml -i interp.py
datum model run -m model-1
datum model run -m model-2

# merge
datum --loglevel debug merge ./test_project-inference.1 ./test_project-inference.2

# compare
datum quality -p test_project-inference.1 --inference-path test_project-inference.2

# just show the boxes
datum project transform -p merged -t boxes_to_masks
datum project export -p undefined-boxes_to_masks/ -f voc_segmentation -e '/item/annotation' --filter-mode i+a -- --save-images

How has this been tested?

Manual test, unit tests.

Checklist

• I submit my changes into the develop branch
• I have added description of my changes into CHANGELOG file
• I have updated the documentation accordingly
• I have added tests to cover my changes
• I have linked related issues (read github docs)
• I have increased versions of npm packages if it is necessary (cvat-canvas,
  cvat-core, cvat-data and cvat-ui)

License

• I submit my code changes under the same MIT License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below)

# Copyright (C) 2020 Intel Corporation
#
# SPDX-License-Identifier: MIT

[coveralls]

Pull Request Test Coverage Report for Build 6926

• 660 of 792 (83.33%) changed or added relevant lines in 14 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.8%) to 69.729%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
datumaro/datumaro/components/algorithms/rise.py	2	4	50.0%
datumaro/datumaro/cli/util/init.py	2	17	11.76%
datumaro/datumaro/components/operations.py	505	538	93.87%
datumaro/datumaro/util/annotation_util.py	109	142	76.76%
datumaro/datumaro/cli/commands/merge.py	12	61	19.67%

Totals
Change from base Build 6850:	0.8%
Covered Lines:	11722
Relevant Lines:	16356

---

💛 - Coveralls

[zhiltsov-max]

@azhavoro, @nmanovic, here is something you can test (in the description).

[nmanovic]

Here is an overview of what got changed by this pull request:

Issues
======
- Added 1
           

Complexity increasing per file
==============================
- datumaro/datumaro/util/annotation_tools.py  1
- datumaro/datumaro/cli/contexts/project/diff.py  1
- datumaro/datumaro/cli/commands/quality.py  7
- datumaro/datumaro/cli/commands/merge.py  7
         

Complexity decreasing per file
==============================
+ datumaro/datumaro/components/project.py  -3
         

See the complete overview on Codacy

[nmanovic]

Codacy found an issue: expected an indented block (, line 95)

[azhavoro]

@zhiltsov-max Could you please update PR? There are many indentation errors in datumaro/datumaro/components/operations.py.

[zhiltsov-max]

Finished.

[nmanovic]

@zhiltsov-max , need to fix merge conflicts

[nmanovic]

LGTM