JLpy Utils Package

Custom modules/classes/methods for various data science, computer vision, and machine learning operations in python

Dependancies

• General libraries distributed with Anaconda (pandas, numpy, sklearn, scipy, matplotlib, etc.)
• image/video analysis:
  • cv2 (pip install opencv-python)
• ML_models sub-package dependancies:
  • tensorflow or tensorflow-gpu
  • dill

Installing & Importing

In CLI:

$ pip install -upgrade JLpy_utils_package

After this, the package can be imported into jupyter notebook or python in general via the comman: import JLpy_utils_package as JLutils

Modules Overview

There are several modules in this package:

JLutils.summary_tables
JLutils.plot
JLutils.img
JLutils.video
JLutils.ML_models

JLutils.summary_tables and JLutils.plot probably aren't that useful for most people, so we won't go into detail on them here, but feel free to check them out if you're curious.

JLutils.img

The JLutils.img module contains a number of functions related to image analysis, most of which wrap SciKit image functions in some way. The most interesting functions/classes are the JLutils.img.auto_crop.... and JLutils.img.decompose_video_to_img().

The auto_crop class allows you to automatically crop an image using countours via the use_countours method, which essentially wraps the function skimage.measure.find_contours function. Alternatively, the use_edges method provides cropping based on the skimage.feature.canny function. Generally, I find the use_edges runs faster and gives more intuitive autocropping results.

The decompose_video_to_img() is fairly self explanatory and basically uses cv2 to pull out and save all the frames from a video.

JLutils.video

...

JLutils.kaggle

This module contains functions for interacting with kaggle. The simplest function is:

JLutils.kaggle.competition_download_files(competition)

where competition is the competition name, such as "home-credit-default-risk"