Open juntao opened 1 year ago
hey @juntao interested in this issue under LFX. I have programming experience in c++ and rust. Will go through the sources you have mentioned and familiarize myself with them. Would be great if you could mention some additional skills that might be required to do this seamlessly.
hi @aashi-ihsaa
Thank you for your interest! Please see the updated description for two pre-tests -- they will help you evaluate the skills needed for this project. I would strongly encourage you to take those pre-tests and apply.
cheers Michael
sure michael .. will take those pre tests. thanks a ton
Motivation
The mediapipe-rs project provides a Rust SDK to support mediapipe AI models. The SDK provides utility functions to pre-process application data (such as images, audio and video) into TFLite / PyTorch formats, and convert the inference results back into application data. In order to accomplish this, the mediapipe-rs project has made extensive use of the ffmpeg library. It compiles ffmpeg to Wasm and then builds it together with the application binary.
However, the issue with this approach is that those Wasm-compiled ffmpeg functions are slow. We believe a better approach is to create a ffmpeg plugin for WasmEdge, and allow Wasm applications to call native ffmpeg functions as host functions.
Expected outcome
The deliverables will be
We expect the deliverable will cover at least 80% of all ffmpeg functions, and have complete unit test and documentation coverage.
Recommended Skills: C/C++, ffmpeg, Rust, WebAssembly
Please familiarize yourself with the following
Apply
Please complete the two pre-tests:
2692
2703
Add your answers under the discussion threads, or email them to me at michael AT secondstate.io
Then, you can apply here: https://mentorship.lfx.linuxfoundation.org/project/41a8bf79-a903-4c03-afb7-256fd373c0b0
Please refer to the timeline of LFX mentorship 2023 03-Sep-Nov. Since we have to process the internal review and decide, the final date for valid submissions will be Aug 27, 5:00 PM PDT.