Sorry, you need to enable JavaScript to visit this website.

Projects that we maintain:


Projects where we actively participate:

  • Sound open firmware
  • FFmpeg vaapi and filters
  • Gstreamer plugins
  • LIbxcam

List of Project Ideas:


  • Fuzzing SOF IPC
       Description:Fuzz the SOF IPC input to make the SOF FW more resilient.There are a few open-source fuzzing tools available such as AFL/Peach/oss-fuzz. The scope of this project involves leveraging the input generation in the open-source fuzzers and integrating it with the SOF fuzzer test harness and QEMU to fuzz the code that parses the topology file and sends the IPCs to the DSP.
       Skill Required: C, Python, Bash, or other scripts language.
       Optional Skills:Familiar with ALSA topology (
       Hardware required: SOF support Intel PC
       Possible mentors: Ranjani Sridharan (, Curtis Malainey (



  • FFmpeg DNN native conv2d layer optimization.
       Description: FFmpeg DNN (deep neural network) module supports the dnn-based filters, it has two backends, one backend is TensorFlow which invokes TensorFlow C libarary for model loading and inference, the other backend is native. The native backend is a CPU fallback option when the system does not support TensorFlow, and so we can’t introduce 3rd party library for native mode. The native mode is still in early development stage and the performance has not been tuned yet. This project focuses on the native conv2d layer optimization with c/asm on Intel CPUs.
        Difficulty: Medium
        Skill Required: C, X86 ASM (AVX2, optional AVX512), DNN knowledge, git
        Hardware requirement: Intel CPU with AVX2/AVX512
        possible mentor: Guo, Yejun (])



  • General depthmap based on 360 dual/stereo camera (eg. Kandao).
       Description: To generate depth map based on 360 dual/stereo camera. 
       Difficulty: Medium
       Skill Required: C/C++
       Optional Skills: OpenCL/OpenCV/Image processing algorithm.
       Hardware Requirement: Intel based PC
       Possible mentor: Zong, Wei (


  • Enable HDR10+/HDR10/HLG based on different exposure images.
       Description: To investigate HDR algorithms based on 2 or 3 Low, (mid), long exposure images into one clear image. Enable the HDR feature into libxcam ( Performance improvements based on Intel CPU/GPU also need to be considered.
       Difficulty: Medium
       Skill Required: C/C++/OpenCL
       Optional Skills: OpenCL/OpenCV/Image processing algorithm.
       Hardware Requirement: Intel Skylake+ based PC
       Possible mentor: Zong, Wei (


  • Title: Add face anti-spoofing 3D Mask function for libxcam
      Description: Design and implement an face anti-spoofing solution by using DNN technique and Intel RealSense camera. Add related APIs into libxcam. Cook a sample program to use anti-spoofing API.
      Difficulty: Medium
      Skill Required: C/C++/python
      Optional Skills: OpenCL/OpenCV/Image processing algorithm.
      Requirement: Intel Skylake+ based PC, Intel RealSense camera
      Possible mentor: Wu, Zhiwen ( Zong, Wei (



  • Enable vaapi based hw decoder on gst-libav
      Description:Gst-libav is an important component in gstreamer. Gstreamer use it to decode/encode almost all video formats in world. However, after many years development, gst-libav still can’t support hw codec. Let us identify the gap and provide necessary patch to fill this gap. The student need modify the gst-libav/gstreamer code to enable hw decoder in gst-libav. The implementation must not copy memory from decoder to renderer(zero copy)
      Difficulty:  Hard
      Skill Required: C, gstreamer
      Optional Skills: git
      Hardware requirement: Intel CPU with integrated GPU since Haswell
      Possible Mentor:
          Xu Guangxin (,  Xiang, Haihao(