Blog: GSoC’23 , Red Hen Lab
Hello there!
My name is Saksham Gautam and I am currently a third year undergraduate at Institute of Engineering and Technology, Lucknow. This summer, as part of Google Summer of Code 2023, I will be contributing to Red Hen Lab.
Throughout the project, I’ll be sharing updates and insights on this blog.
Outline
*Will get updated as the project progresses*
- Community Bonding period (4th May — 29th May)
- Examination break(29th May — 9th June)
*Coding Officially begins*
- Week 1(9th June— 14th June)
- Week 2(14th June— 21st June)
- Week 3 (21st June — 28th June)
- Week 4(28th June — 5th July)
- Week 5(5th July — 12th July)
- Week 6(12th July — 19th July)
- Week 7(19th July — 26th July) (Mid term evaluation begins)
- Week 8(26th July — 2nd August)
- Week 9(2nd August — 9th August)
- Week 10(9th August — 16th August)
- Week 11(16th August — 23rd August)
- Week 12(23rd August — 30th August)
- Week 13(30th August — 6th September)
- Week 14(6th September — 13th September)
Mentors
- Daniel Alcaraz Carrión
- Mark Turner
- Peter Uhrig
Abstract
The Red Hen Anonymizer is a software that uses deep learning and signal processing techniques to anonymize audio-visual data. The proposed project aims to expand the software’s capabilities by adding two new features for visual data anonymization, namely Level Wise Anonymization and Anonymization using StarGANv2. Additionally, the project proposes to add a new feature for audio data anonymization, namely Anonymization According to the Anonymized Face, while retaining the existing features.
The new algorithms are based on deep learning and require GPU power for training. The objective is to create a versatile and robust software, with a user-friendly interface. The final deliverable will be a complete anonymization engine integrated into a web application, providing users with a convenient one-click anonymization options that meets their specific needs.
Community Bonding Period
Week 1
- Setup CWRU HPC account
- Study thoroughly Techne Page on CWRU HPC
- Learn use of screen
Week 2
- Run RHA as it is present in Yash Khasbage’s directory(yck5). rha.py file requires a temporary directory to be made to store intermediate results. But sadly I don’t write access to yck5’s directory. Solution-: Copy rha.py file to my gallina home and form a pipeline to yck5 gallina directory where the remaining files are present.
- Install anaconda3, sox and its dependencies (http://akshayc.com/blog/build-sox-from-scratch/) in my home directory and set the path variables accordingly. The main challenge is to install sox from scratch as we are not a super user(in short we cannot use sudo).
- Make a conda environment and install all the dependencies of RHA. (https://yashkhasbage25.github.io/AnonymizingAudioVisualData/2021/08/14/installation.html).
Week 3
- Prioritized the goals of the project. The ultimate aim is to reduce the flickers due to face swapping.
- Discussed a probable extension of the project with the mentors due to end semester examinations.
- Project extension approved! Project converted to a 16 weeks project.