Julian Tanke
Julian Tanke
It seems like left and right are flipped in your 3D representation, e.g. the following code https://github.com/una-dinosauria/human-motion-prediction/blob/c9a2774fe59bb78c6ec6d04e15f2cb7243c4348c/src/viz.py#L25 seems to conflict with the human 3.6m matlab code ```matlab switch part case...
* MOTP * MOTA For that we need to reproduce their detection "ground-truth", e.g. run over the gt and the detection data, find the bb's with IoU > 0.5 and...
When loading broken videos (e.g. empty file) we should gracefully fail and show a error message
I am a bit confused about this two lines: https://github.com/cr7anand/neural_temporal_models/blob/0b72be14e1fba0ed681322994e09116496ff19e8/metrics.py#L54 and https://github.com/cr7anand/neural_temporal_models/blob/0b72be14e1fba0ed681322994e09116496ff19e8/metrics.py#L77 I think ``` gt_fourier_coeffs = np.zeros(euler_gt_sequences.shape) ``` should actually be ``` gt_fourier_coeffs = np.zeros(euler_gt_sequences.shape, dtype=np.complex) ``` as you...
I was wondering at what framerate your CMU data is sampled. Human3.6M is sampled at `50Hz` and downsampled to `25Hz` for training/evaluation but the original CMU dataset is sampled at...