Problems launching kitti example

[DaddyWesker]

Hello.

I've been trying for a while to launch your example with kitti using command
python evaluate_depth.py --data_path /home/daddywesker/datasets/kitti/ --load_weights_folder ./ --post_process --save_pred_disp --eval_mono
UNfortunately, i'm getting

`TypeError: init() got an unexpected keyword argument 'do_not_load_pretrained'

AS i see in networks/resnet_encoder there is no do_not_load_pretrained in init for this network. SO, probably, either this 'do_not_load_pretrained' was deleted in the new version or wasn't added. If i'm changing line from
encoder = networks.ResnetEncoder(cfg, maskrcnn_path, do_not_load_pretrained=True)
to encoder = networks.ResnetEncoder(cfg, maskrcnn_path)

i'm getting

Traceback (most recent call last):
  File "/home/daddywesker/TestingRGB/SimVODIS/evaluate_depth.py", line 237, in <module>
    evaluate(options.parse())
  File "/home/daddywesker/TestingRGB/SimVODIS/evaluate_depth.py", line 96, in evaluate
    encoder = networks.ResnetEncoder(cfg, maskrcnn_path)#, do_not_load_pretrained=True)
  File "/home/daddywesker/TestingRGB/SimVODIS/networks/resnet_encoder.py", line 62, in __init__
    _ = self.checkpointer.load(pretrained_model_path, convert=True)
  File "/home/daddywesker/TestingRGB/SimVODIS/maskrcnn_benchmark/utils/checkpoint.py", line 67, in load
    self._load_model(checkpoint)
  File "/home/daddywesker/TestingRGB/SimVODIS/maskrcnn_benchmark/utils/checkpoint.py", line 111, in _load_model
    load_state_dict(self.model, checkpoint.pop("model"))
  File "/home/daddywesker/TestingRGB/SimVODIS/maskrcnn_benchmark/utils/model_serialization.py", line 77, in load_state_dict
    align_and_update_state_dicts(model_state_dict, loaded_state_dict)
  File "/home/daddywesker/TestingRGB/SimVODIS/maskrcnn_benchmark/utils/model_serialization.py", line 35, in align_and_update_state_dicts
    max_match_size, idxs = match_matrix.max(1)
RuntimeError: cannot perform reduction function max on tensor with no elements because the operation does not have an identity

so, probably, there should be some replacement... Any advice?

[DaddyWesker]

Well, error with cannot perform reduction function is related to line
checkpoint = self._convert_checkpoint(checkpoint_old)
in the file maskrcnn_benchmark/utils/checkpoint.py. After this operation checkpoint is empty. I've set convert to False but not sure if it is a good solution...

[DaddyWesker]

With this "solution" of mine program worked till the end and gave me some info

 Mono evaluation - using median scaling
 Scaling ratios | med: 34.752 | std: 0.026

   abs_rel |   sq_rel |     rmse | rmse_log |       a1 |       a2 |       a3 | 
&   0.079  &   0.270  &   3.028  &   0.107  &   0.946  &   0.998  &   1.000  \\

-> Done!

Process finished with exit code 0

So, how to get semantic images as in here https://github.com/Uehwan/SimVODIS#qualitative-results ?