Open yy19931029 opened 4 years ago
Hi,
I never got round to packaging the CCNF eye model training code, but it is the same approach as for the other parts of the face, just at a different scale.
The reason they are done at a higher scale is because the training was done on synthetic data I could not use a lower scale due to the synthetic data modeling only the "zoomed in" eye region and lower scale would include capturing much more of the face region not modeled by synthetics.
First of all, I would like to thank you for your work! I wanted to train my eye model, but you did not provide the training code. your ccnf_eye model in model\model_eye\patch_experts\ ccnf_patches_1.50_synth_lid only have scale in 1.0 and 1.5. when my face small, whether it will lead to the poor detection effect , and whether the model scale in 0.25 0.3and 0.5 can be provided. Thank you very much!