This repository contain all util.

Get at https://pypi.org/project/dpsutil/

pip install dpsutil

Ver 1.0.0:

• Compression -> DONE
• KafkaWrapper -> DONE
• RedisWrapper -> DONE
• Vector -> DONE
• Media -> DONE
• Computer Vision -> DONE
• Distance -> DONE
• Hashing -> DONE

Ver 1.0.1:

• Added image_info -> Get (format, width, height) of image without decoding.
• Upgrade imdecode -> Auto decode image but need not provide encode_type.

Ver 1.0.2:

• Change 'numpy.pool' to 'vector.pool'
• Fix bugs

Ver 1.0.3:

• Compression:
  • Support list compression: compress_list, decompres_list
• Distance:
  • Change distance.function -> vector.distance

Ver 1.1.0:

• Added AttrDict, FixedDict, FixedTypeDict

Ver 1.1.1-1.1.7:

• Fix bugs
• Added UniqueTypeDict
• Fix bugs
• Changed module fixdict -> defaultdict
• Changed FixedDict -> DefaultDict
• Changed FixedTypeDict -> DefaultTypeDict
• Fix bugs of attrdict.AttrDict, attrdict.UniqueTypeDict, attrdict.DefaultDict
• Fix bugs of compress_list, decompress_list

Ver 1.1.8:

• Added Environment

Ver 1.1.9:

• Added Environment.to_lower -> Useful when extracting to kwarg.

Ver 1.1.10-1.1.15:

• Fix bugs attrdict
• Changed hashfunc default from sha1 -> md5
• Fix bugs, casting type when set value.

Ver 1.1.16-1.1.18:

• Support call of DefaultDict, which update data from buffer.
• Fix bugs Environment

Ver 1.1.19:

• Added pre-define output array of decompress_ndarray
• Support add new key environment.

• Numpy Pool -> Cache Algorithm in RAM
• Sort -> implement natsort - https://github.com/SethMMorton/natsort
• CV -> Find more util
• Find & Add more functions

• Support type: ndnumpy, bytes
• Compress by blosc. It support multi compressor and multi-thread.

• Wrapping Consumer and Producer with default setting and security.

• Wrapping Redis Connector with default setting and security.

• Implemented numpy.memmap with High Performance and control memory IO.

Implemented OpenCV with:

• TurboJPEG (https://github.com/lilohuang/PyTurboJPEG)
• FFmpeg(https://github.com/kkroening/ffmpeg-python)

To: improve read & write (image, video) IO speed.

• Faster than OpenCV:

  • 2-6x in JPEG
  • 1.1x with others.

• Added some function which used frequently.

• More info: find in dpsutil.media

• Added Face Align with five landmark.

• Added cosine_similarity
• Added cosine
• Added euclidean_distance
• Added convert distance functions

*Note: all function execute in numpy.

• AttrDict: will help you get value of key via attributes. Implement attrdict.AttrDict
• FixedDict: help cover your dict with (keys, values) that was defined before. Implement from AttrDict
• FixedTypeDict: help cover your dict when set item. Implement from FixedDict.
• UniqueTypeDict: dict only access one type for all element.

• Environment: Auto pair environment parameters fit with default, which provided before.

• Hashing
• Sort

Firstly, you need install scikit

$ pip install scikit-build

After that, follow instuction to install Cmake:

https://cliutils.gitlab.io/modern-cmake/chapters/intro/installing.html

Python

$ pip install cmake>=3.12   

Find FFmppeg lib at:

https://www.ffmpeg.org/download.html

Linux:

sudo apt-get install ffmpeg

Find FFmppeg lib at:

https://libjpeg-turbo.org/Documentation/OfficialBinaries

Linux:

$ sudo apt install libturbojpeg

• Make sure your Redis or Kafka server started.
• Make sure correct username & password.

SSL development libraries have to be installed

https://github.com/openssl/openssl#build-and-install

Ubuntu:

$ apt-get install libssl-dev libffi-dev

Develop by: TinDang
OS: Ubuntu Bionic 18.4
Python: 3.6.9 stable