Changing stuff to accomondate the new DF3D API

[sibocw]

I think this addresses probably most of the incompatibilities between df3dpp and the new df3d API.

Exactly what I did / extent of testing: I tried to run Florian's behavior data processing on the data used for the DF3D paper. The DF3D outputs were regenerated in February using the new DF3D API. (I believe --- all output files that begin with df3d_result_*.pkl are generated with the new version, right?) The df3dpp package is also installed with separate (new) nely-df3d and nely-pyba packages. It seems to be working now, at least to me:

Can the next user review the changes, and if everything looks good, merge please?

[sibocw]

Trickiest changes:

• deepfly.cv_util.triangulate_linear is moved to pyba.util.triangulate, and the arguments changed
• pyba.Camera.project now strictly accepts a (TxJx3) array. For per-joint, per-frame calibration, just set T=1, J=1.

[sibocw]

Works on all data in the DF3D paper except 180918_MDN_CsCh_Fly3_004_SG1. Probably something to do with too many outliers (?) --- not super important for now.

Traceback (most recent call last):
  File "/home/sibwang/Projects/drosophila-imitation-learning/scripts/preprocessing/gunel2019_df3d_df3dpp.py", line 84, in <module>
    processed_dir / '180918_MDN_CsCh_Fly3_004_SG1/post_processed.pkl'
  File "/home/sibwang/Projects/drosophila-imitation-learning/scripts/preprocessing/gunel2019_df3d_df3dpp.py", line 31, in process_trial
    calculate_3d=True, outlier_correction=True)
  File "/home/sibwang/Projects/df3dPostProcessing/df3dPostProcessing/df3dPostProcessing.py", line 229, in __init__
    outlier_correction=outlier_correction)
  File "/home/sibwang/Projects/df3dPostProcessing/df3dPostProcessing/df3dPostProcessing.py", line 361, in load_data
    data, exp, outlier_correction)
  File "/home/sibwang/Projects/df3dPostProcessing/df3dPostProcessing/df3dPostProcessing.py", line 390, in triangulate_2d
    camNet = correct_outliers(camNet)
  File "/home/sibwang/Projects/df3dPostProcessing/df3dPostProcessing/utils/utils_outliers.py", line 157, in correct_outliers
    camNet = interpolate_remaining_outliers(camNet, outliers)
  File "/home/sibwang/Projects/df3dPostProcessing/df3dPostProcessing/utils/utils_outliers.py", line 90, in interpolate_remaining_outliers
    non_outlier_frames, non_outlier_values, axis=0)
  File "/home/sibwang/anaconda3/envs/flynet1030b/lib/python3.6/site-packages/scipy/interpolate/_cubic.py", line 100, in __init__
    dk = self._find_derivatives(xp, yp)
  File "/home/sibwang/anaconda3/envs/flynet1030b/lib/python3.6/site-packages/scipy/interpolate/_cubic.py", line 173, in _find_derivatives
    dk[0] = PchipInterpolator._edge_case(hk[0], hk[1], mk[0], mk[1])
IndexError: index 0 is out of bounds for axis 0 with size 0

[sibocw]

Copying potential next users: @jasper-tms @jonasfbraun