Frame Snipper

A Python program that trims off undesired of clips that matches in visual similarity with the undesired templates

Getting Started

These instructions will get you a copy of the project up and running on your local machine.

Prerequisites

• Python 3

Dependencies

• numpy
• openCV
• Pillow

Installing

Run the following from the command line

pip install -r requirements.txt

Quick Start

Example: python snip.py -i input_folder -o output_folder template_file.csv

The content of the template_file.csv is in the form of

FILE_NAME, START_TIMESTAMP, END_TIMESTAMP, START_RANGE, END_RANGE

Where FILE_NAME specifies the video file to act as the template

START_TIMESTAMP/END_TIMESTAMP is the start/end of the segment to excise

START_RANGE/END_RANGE is the region where the program should search.

Timestamps can be specified in seconds or MM:SS.

Note that negative values is allowed for the time stamps, to indicate duration from the end of the clip.

For example

Ep1.mp4,1:49,3:08,0,300
Ep1.mp4,22:29,23:35,-300,0

If no input folder is specified, it will default to using input as the the folder.

If no output folder is specified, it will default to using output as the the folder.

If you do not wish to trim the clip from the template file, add the -s or --skip-templates option.

To check if the timestamps specified is correct, the -f or --viewframe argument can be added to the program.

This will display the start/end frames as specified in the template

Example Workflow

Firstly, we procure a set of video clips with a common segment that we wish to excise.

We manually determine the timestamp of the portion that we wish to excise for the first clip. Use python snip.py -i INPUT_FOLDER -f TEMPLATE.csv to preview the starting/ending of the segments that will be excised.

Once we determined that the template is acceptable, use python snip.py -i INPUT_FOLDER -o OUTPUT_FOLDER TEMPLATE.csv to start the excising process.

Details

• The starting/ending frame of the segment which we want to excise remain relatively unchange for a period of time, or else the program may be unable to find this frame in the target clip (unless the granularity is set to 0). Thus, ideally, our clip should start on a frame that does not rapidly change, ie it does not cut to a different scene.

• The starting frame that we excise should be unique enough that the program may not excise based on a false positive. Thus, it is not advisable to excise based on a complete black/white frame.

• The program determines the similarity using histogram matching of the HSV channels.

• The program assumes that the duration of the excised clips are constant for all the episodes. Meaning it does not search for similarity of the end of the template.

Authors

• 0WN3D - Initial work - 0WN463

License

This project is licensed under the MIT License