dreamlessx commited on
Commit
20a1910
·
verified ·
1 Parent(s): 1654955

Update README with features section and photo tips

Browse files
Files changed (1) hide show
  1. README.md +20 -5
README.md CHANGED
@@ -16,14 +16,22 @@ tags:
16
  - landmarks
17
  - thin-plate-spline
18
  - surgery-simulation
 
 
19
  ---
20
 
21
  # LandmarkDiff
22
 
23
  Anatomically-conditioned facial surgery outcome prediction from standard clinical photography.
24
 
25
- Upload a face photo, select a surgical procedure, adjust intensity, and see the predicted outcome
26
- in real time using thin-plate spline warping on CPU.
 
 
 
 
 
 
27
 
28
  ## Supported Procedures
29
 
@@ -39,12 +47,19 @@ in real time using thin-plate spline warping on CPU.
39
  ## How It Works
40
 
41
  1. **MediaPipe landmarks** -- 478-point facial mesh extraction
42
- 2. **Anatomical displacement** -- procedure-specific landmark shifts (intensity 0-100)
43
  3. **TPS deformation** -- thin-plate spline warps the image smoothly
44
- 4. **Masked compositing** -- blends the surgical region back into the original
45
 
46
  GPU modes (ControlNet, img2img) with photorealistic rendering are available in the full package.
47
 
 
 
 
 
 
 
 
48
  ## Links
49
 
50
  - [GitHub](https://github.com/dreamlessx/LandmarkDiff-public)
@@ -52,4 +67,4 @@ GPU modes (ControlNet, img2img) with photorealistic rendering are available in t
52
  - [Wiki](https://github.com/dreamlessx/LandmarkDiff-public/wiki)
53
  - [Discussions](https://github.com/dreamlessx/LandmarkDiff-public/discussions)
54
 
55
- **Version:** v0.2.2
 
16
  - landmarks
17
  - thin-plate-spline
18
  - surgery-simulation
19
+ - facial-analysis
20
+ - symmetry
21
  ---
22
 
23
  # LandmarkDiff
24
 
25
  Anatomically-conditioned facial surgery outcome prediction from standard clinical photography.
26
 
27
+ Upload a face photo, select a surgical procedure, adjust intensity, and see the predicted outcome in real time using thin-plate spline warping on CPU.
28
+
29
+ ## Features
30
+
31
+ - **Single Procedure**: Predict the outcome of one procedure at a chosen intensity
32
+ - **Compare All**: See all six procedures side by side at the same intensity
33
+ - **Intensity Sweep**: View a single procedure from 0% to 100% in six steps
34
+ - **Symmetry Analysis**: Measure bilateral facial symmetry across five regions, with pre/post comparison
35
 
36
  ## Supported Procedures
37
 
 
47
  ## How It Works
48
 
49
  1. **MediaPipe landmarks** -- 478-point facial mesh extraction
50
+ 2. **Anatomical displacement** -- procedure-specific landmark shifts scaled by intensity (0-100)
51
  3. **TPS deformation** -- thin-plate spline warps the image smoothly
52
+ 4. **Masked compositing** -- blends the surgical region back into the original photo
53
 
54
  GPU modes (ControlNet, img2img) with photorealistic rendering are available in the full package.
55
 
56
+ ## Photo Tips
57
+
58
+ - Use a front-facing, well-lit photo with a neutral expression
59
+ - Remove glasses, hats, or anything covering the face
60
+ - Make sure only one face is clearly visible
61
+ - At least 256x256 resolution recommended
62
+
63
  ## Links
64
 
65
  - [GitHub](https://github.com/dreamlessx/LandmarkDiff-public)
 
67
  - [Wiki](https://github.com/dreamlessx/LandmarkDiff-public/wiki)
68
  - [Discussions](https://github.com/dreamlessx/LandmarkDiff-public/discussions)
69
 
70
+ **Version:** v0.2.2 | **License:** MIT | For research and educational purposes only