Loic Landrieu

Hi, I retrained a classifier from scratch using: ``` CUDA_VISIBLE_DEVICES=0 python3.6 learning/main.py --dataset sema3d --SEMA3D_PATH $SEMA3D_PATH --db_test_name testred --db_train_name trainval --epochs 500 --lr_steps '[350, 400, 450]' --test_nth_epoch 1000 --model_config 'gru_10,f_8'...

It will come! Early 2020.

Right, an error happens if the size of the file is a multiple of the batch size. Good catch! It is fixed in the latest commit. Also please use the...

1. you need to recompile libply_c.so 2. Yes, that's expected. I'll look at how to shut down this warning later. 3. I wouldn't worry about it, 5 iterations of cut...

Hi, Can you print the size of `vertices` just before the bug? Also priniting the batch number : ``` print("%d" % (i_batch)) ``` With `i_batch` a counter incremented at the...

Hi, can you add the following line 238 of `/partition/provider.py`, just before `xyz_full = ...` ``` print(vertices.shape) ``` and report the log after: ``` 7 / 16---> test_full creating the...

Hi, the test labels for testred and testfull are withheld by semantic3d. If you want to evaluate your performance you should do a submission on their website. To use the...

Is this during training or inference?

Ah yes. The hard cutoff is the maximum number of "valid" superpoints to embed with the CloudEmbedder. A superpoint is valid if it has more than `args.ptn_minpts` points. Indeed, superpoints...

replacing: ``` G = k_big_enough(G, args.ptn_minpts, args.spg_augm_hardcutoff) ``` by ``` G = G.subgraph(range(args.spg_augm_hardcutoff)) ``` should work.