uniform landmark distribution vs. landmark identification

[martin-raden]

test the performance when landmarks are not computed but just uniformly sampled/assigned over the whole curve