Peekingduck

**Features**

- Add Python API mode by allowing configs to be modified at node level
- Simplify calling of config attributes as a consequence of the above change

python
Example using input.live's __init__ method

Previously, configs had to be extracted from the "config" dictionary
mirror_image = config['mirror_image']

Now, the Abstract Node takes care of this, and each attribute can be called directly
self.mirror_image


- Improve error handling of `input`, `model`, and `output` nodes

**Deprecations and Removals**

- Rename all `heuristic` nodes as `dabble` nodes

**Improved Documentation**

- Shorten [README.md](http://readme.md) by moving most guides into readthedocs page ([https://peekingduck.readthedocs.io/en/stable/](https://peekingduck.readthedocs.io/en/stable/))
- Include bibliography, Python API and CV model benchmark sections

**Dependencies**

- Move CI/CD dependencies out of `requirements.txt` into `cicd_requirements.txt`

**User Interface**

- Update colours of bboxes, keypoints, edges, fonts to maintain consistency of PeekingDuck colour scheme
- Improve visibility of displayed text by adding translucent/opaque background behind

**Features**

- Add `draw.legend` node for drawing output of `heuristic.bbox_count`, `heuristic.zone_count`, `heuristic.fps` in a legend
- Add `draw.image_processor` node to allow changing of image brightness and contrast
- Add `heuristic.fps` node for tracking FPS
- Add feature to modify node configurations from CLI by using `--node_config` flag
- Add feature to show version of PeekingDuck installed from CLI by using `--version` flag

**Deprecations and Removals**

- Deprecate `draw.bbox_count` node
- Deprecate `draw.zone_count` node
- Deprecate `draw.fps` node

**Improved Documentation**

- Standardise node docstring formatting for readthedocs page ([https://peekingduck.readthedocs.io/en/stable/](https://peekingduck.readthedocs.io/en/stable/))
- Update downloadable demo video in readme to show ducks instead of athletes

**Bug Fixes:**
- Fix custom node creation for Windows users

**Process**

- Update code and docs on modifying node configurations in `run_config.yml` to follow typical yaml convention without requiring a `-` in front
jsx
Updated
- input.live:
resize:
do_resizing: True

Previous
- input.live:
- resize:
do_resizing: True


**Features**

- Add `model.hrnet` node for pose estimation
- Add `output.csv_writer` node
- Allow folder name for custom node source code to be customised using `-custom_folder_name` flag
- Allow video resolution to be changed in configs for `input.live` and `input.recorded`
- Allow live feed from `input.live` to be saved as a recorded video after inferencing
- Add `["all"]` input type that allows nodes to access all variables in data pool if used
- Add `["pipeline_end"]` output type that returns True when the pipeline has ended for `input.live`, `input.recorded` and `output.screen` nodes
- Add `["none"]` input/output type for nodes that do not require any input/output
- Show percentage progress of inference on recorded videos using `input.recorded`
- Show updates of number of processed frames using `input.live`
- Append timestamp to output filenames

**Deprecations and Removals**

- Remove `["end"]` input/output type - this is replaced by `["pipeline_end"]`
- Remove `["source"]` input type - this is replaced by `["none"]`
- `draw` nodes return `["none"]` instead of `["img"]` types

**Bug Fixes**

- Update nested configs in `run_config.yml` without overwriting key-value pairs of the same parent

Minor bug fixes to enable python 3.8 support.
- loosen numpy version required `numpy >= 1.17.3`
- loosen pyyaml version required `pyyaml >= 5.3`