dpsailab-util 1.4.2

This repository contain all util for DPS AI Computer Vision's member

DPS_Util

This repository contain all util.

Install:

Get at https://pypi.org/project/dpsailab-util/

pip install dpsailab-util

Change log:

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 1.1.9-1.1.18:

• Added Environment.to_lower -> Useful when extracting to kwarg.
• Fix bugs attrdict
• Changed hashfunc default from sha1 -> md5
• 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.

Ver 1.1.20-1.1.21:

• Edit logger
• Fix bugs vector.pool
• Changed VectorPool base on numpy.mmap -> VectorPoolMMap
• Added VectorPool base on numpy.ndarray
• Restructure VectorPoolBase -> Speed up, fix bugs
• Added VectorPoolBase.insert

Ver 1.1.22:

• Add document of compression, media module.
• Fix bugs image
• Change media.image.scale -> media.image.zoom

Ver 1.1.23-24:

• Fix bugs log
• Fix bugs VectorPoolMMap.MIN_SIZE

Ver 1.1.25-26:

• Fix bugs attrdict
• Added add, remove method attrdict
• Rename variable of pool.VectorPoolBase

Ver 1.1.27:

• Change namespace attrdict.UniqueTypeDict -> attrdict.TypedDict
• Restructure attdict.AttrDict, attdict.DefaultDict, attrdict.TypedDict
• Support annotations alias
• Speed up attrdict
• Support initial attrdict with iterable, generator.
• Support to_buffer method with optimize compression.
• Support recreate dict with from_buffer.
• Lightweight than pickle.dumps

Ver 1.2.0:

• CV:
  • Added Face Aligner
  • Transforms image methods
• media.image:
  • Added scale, crop_center, flip
  • Fix bugs
• media.video: support FPS report.
• triangle (new)

Ver 1.3.0:

• Support decorator attrdict
• Support decorator environment
• Implemented Confluent-Kafka with:
  • Iterator method
  • Close event.
  • Auto_commit
  • Optimize configs for ThroughOut
  • Fix bug KafkaError

Todo:

• Implement Configs librdkafka
• Sort -> implement natsort - https://github.com/SethMMorton/natsort

---

Features

---

Compression Lossless:

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

KafkaWrapper:

• Wrapping Consumer and Producer with default setting and security.

RedisWrapper

• Wrapping Redis Connector with default setting and security.

Vector Pool:

VectorPoolBase handle numpy.ndarray that save a lot of RAM by re-use existed memory space.

Easier than numpy.ndarray. Likely list.

• Reduce RAM
• High Speed IO
• Auto Scale
• Support backup and recovery data in a second.

Media:

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 with JPEG Encoder
  • 1.1x with others.

• Added some function which used frequently.

• More info: find in dpsutil.media

Computer Vision (cv):

• Added Face Align with five landmark.

Distance:

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

*Note: all function execute in numpy.

Attributes dict:

• AttrDict: will help you get value of key via attributes. Implement attrdict.AttrDict
• DefaultDict: help cover your dict with (keys, values) that was defined before. Implement from AttrDict
• TypedDict: help cover your dict when set item. Implement from AttrDict.
• DefaultTypeDict: dict only access one type for all element.
• Support initial attrdict with iterable, generator.
• Support to_buffer method with optimize compression.
• Support recreate dict with from_buffer.
• Lightweight than pickle.dumps

Environement:

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

Other:

• Hashing
• Sort

---

Issue:

---

Cmake error during install blosc

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   

---

Not found 'FFmpeg':

Find FFmppeg lib at:

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

Linux:

sudo apt-get install ffmpeg

---

Not found 'libjpeg-turbo':

Find FFmppeg lib at:

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

Linux:

$ sudo apt install libturbojpeg

---

Not found Redis or Kafka server:

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

Not found 'Could NOT find OpenSSL':

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 Maintain by: CanhTran OS: Ubuntu Bionic 18.04
Python: 3.6.9 stable