csiro-robotics / incloud Goto Github PK

When I tried to run the following code, one error occurred.
python training/train.py --config config/protocols/4-step.yaml --train_environment pickles/Oxford/Oxford_train_queries.pickle

Error info:
Traceback (most recent call last): File "/home/ericxhzou/Code/InCloud/training/train.py", line 44, in <module> trained_model = trainer.train() File "/home/ericxhzou/Code/InCloud/training/trainer.py", line 122, in train self.training_step(batch, positives_mask, negatives_mask) File "/home/ericxhzou/Code/InCloud/training/trainer.py", line 84, in training_step loss_place_rec, num_triplets, non_zero_triplets, embedding_norm = self.loss_fn(embeddings, positives_mask, negatives_mask) File "/home/ericxhzou/Code/InCloud/losses/pr_loss.py", line 80, in __call__ 'num_non_zero_triplets': self.loss_fn.reducer.triplets_past_filter, File "/root/.pyenv/versions/3.10.13/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1269, in __getattr__ raise AttributeError("'{}' object has no attribute '{}'".format( AttributeError: 'AvgNonZeroReducer' object has no attribute 'triplets_past_filter'

Hi, thanks the great work! I would like to know if your other work, GeoAdapt, could also be open-sourced. Thank you very much!

Dear authors,

Thanks for your exciting work.
I am reproducing your results and have noticed that logg3d is not yet implemented. So I take the official code of logg3d to forward the global embeddings(256-dim vectors) and use the default config as the code you open source. Everything seems just fine since the code is successfully running. However, the recall@1 and recall@1% on Oxford and In-house datasets are too low to believe. Do you have any ideas on this issue? You have mentioned in your paper that oxford and In-house lack insufficient overlap between adjacent point clouds. Any influence on the training of logg3d? Hope you can help me solve this issue at your convenience.

Hi authors,

Thanks for your fantastic work!

I have just read your paper and found that in Fig.4 all three PR methods perform better on Oxford after being trained on Riverside, which seems like the opposite trend to catastrophic forgetting. Could you please explain the results?

In the code eval_singlesession.py, should the line 86 elif top1_world_dist > configs.eval.false_positive_thresh: be changed to elif top1_world_dist > false_pos_thresh?

Traceback (most recent call last):
File "training/train_incremental.py", line 79, in
eval_stats = evaluate(new_model, env_idx)
File "/home/ros/deep_learning/Place_recognition/continual_PR/InCloud/eval/evaluate.py", line 26, in evaluate
env_recall_1.append(eval_singlesession(model, d, world_thresh, false_pos_thresh, time_thresh)['Recall@1'])
File "/home/ros/deep_learning/Place_recognition/continual_PR/InCloud/eval/eval_singlesession.py", line 86, in eval_singlesession
elif top1_world_dist > configs.eval.false_positive_thresh:
File "/home/ros/anaconda3/envs/minklocv2/lib/python3.8/site-packages/torchpack/utils/config.py", line 17, in getattr
raise AttributeError(key)
AttributeError: false_positive_thresh