diff options
| author | yum <yum.food.vr@gmail.com> | 2025-07-28 02:16:09 -0700 |
|---|---|---|
| committer | yum <yum.food.vr@gmail.com> | 2025-07-28 02:16:09 -0700 |
| commit | 341ea861c8189cefe8689bd41d8adbe2cd2d87b2 (patch) | |
| tree | 0da0e92cd230a3b371d630eb83bfded99777d733 | |
| parent | 807199c10385eb7b905ae22e41880d671d6f9639 (diff) | |
add gallery
| -rw-r--r-- | Images/fft_blue_noise.png | bin | 0 -> 690932 bytes | |||
| -rw-r--r-- | Images/fft_delta.png | bin | 0 -> 92765 bytes | |||
| -rw-r--r-- | Images/fft_mandrill.png | bin | 0 -> 1353191 bytes | |||
| -rw-r--r-- | README.md | 25 |
4 files changed, 21 insertions, 4 deletions
diff --git a/Images/fft_blue_noise.png b/Images/fft_blue_noise.png Binary files differnew file mode 100644 index 0000000..a8b0a34 --- /dev/null +++ b/Images/fft_blue_noise.png diff --git a/Images/fft_delta.png b/Images/fft_delta.png Binary files differnew file mode 100644 index 0000000..4cf7f45 --- /dev/null +++ b/Images/fft_delta.png diff --git a/Images/fft_mandrill.png b/Images/fft_mandrill.png Binary files differnew file mode 100644 index 0000000..bf94f6a --- /dev/null +++ b/Images/fft_mandrill.png @@ -3,10 +3,6 @@ This is an optimized GPU-based 2D FFT for VRChat. Since it uses RenderTextures, it is only suitable for use in worlds. -Please take note that the FFT is generated in bit-reversed order. You'll have -to bit-reverse your indices when accessing it. See `apply_2d_bit_reversal` in -`gpu_fft.cc` for an example of how to do that. - ### Quick start Clone the repo into your unity project. Open the scene. @@ -35,3 +31,24 @@ compute FFTs with a shorter CRT chain. `DFT_MATRIX` corresponds to `ShaderUniforms.twiddle_factors` in the simulator and `STAGE_TWIDDLES` corresponds to `ShaderUniforms.stage_twiddles`. +### Gallery + +Each of these images shows the following, from left to right: + +1. Original image. +1. FFT stage 0. +2. FFT stage 1. +3. FFT stage 2. +4. FFT stage 3. +5. FFT bit reversal. Numpy reference image is shown above. +6. IFFT stage 0. +7. IFFT stage 1. +8. IFFT stage 2. +9. IFFT stage 3. +10. IFFT bit reversal. The original signal is recovered. + + + + + + |
