How to run train_s1n02_tsdfhist?

I'm trying to run train_s1n02_tsdfhist.

I could not get gentrain_.*_feat_s1n01_r128.oc from training. I got following error message.

[INFO] add 1840 target paths
0
1840
/home/ando/torch/install/bin/luajit: ../common.lua:207: invalid number of paths
stack traceback:
        [C]: in function 'error'
        ../common.lua:207: in function 'addPaths'
        ../common.lua:228: in function 'addTargetPaths'
        modelnet_fusion_common.lua:29: in function 'add_target_paths'
        ../completion_common.lua:365: in function 'get_train_data_loader'
        ../completion_common.lua:383: in function 'get_data_loader'
        ../completion_common.lua:544: in function 'ctf_loop'
        train_s1n02_tsdfhist.lua:29: in main chunk
        [C]: in function 'dofile'
        ...ando/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
        [C]: at 0x00405d50

How to solve this?

Also, let me share several solutions related to this issue. I hope you can correct them if you find my misunderstandings.

create_data.py seems to generate s1n01, but train_s1n02_tsdfhist requires s1n02, so I replaced with s1n01.
I could not find common.json so I added local json = require 'cjson' at the beginning of the file

Hi,

I am having the same issue also. Had several problems that I solved until arriving here (including the two that you mentioned at the bottom). I have the same output than you:

[INFO] add 0 input paths - generated
[INFO] add 1840 target paths
0
1840
/home/lroldoaj/torch/install/bin/luajit: ../common.lua:209: invalid number of paths
stack traceback:
    [C]: in function 'error'
    ../common.lua:209: in function 'addPaths'
    ../common.lua:230: in function 'addTargetPaths'
    modelnet_fusion_common.lua:39: in function 'add_target_paths'
    ../completion_common.lua:372: in function 'get_train_data_loader'
    ../completion_common.lua:390: in function 'get_data_loader'
    ../completion_common.lua:551: in function 'ctf_loop'
    train_s1n02_tsdfhist.lua:35: in main chunk
    [C]: in function 'dofile'
    ...doaj/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
    [C]: at 0x00405d50

Were you able to solve this ?

Btw,

For the problem not finding common.json, what I did was to change: local data = common.json:decode(content) by local data = json.decode(content). Don't forget to import the module by adding local json = require('cjson') at the top of the file.
For the problem about s1n02, the issue is that we have tsdf_hist as input encoding, which not corresponds to the settings we had when we created the data by running the create_data.py script. Changing to s1n01 is the right setting for this input. s1n02 would correspond to the occupancy encoding 'oc'. Check on the create_data.py file:
```
Settings = [
Setting(view_setting_idx=1, encoding='occ', vx_resolutions=sel_vx_resolutions, noise=(0.02, 'n02')),

Setting(view_setting_idx=1, encoding='tsdf_hist', vx_resolutions=sel_vx_resolutions, noise=(0.01, 'n01')),

# GT
Setting(view_setting_idx=0, encoding='tsdf', vx_resolutions=sel_vx_resolutions, noise=(0,'')),
]
```

Would you recommend any particular IDE for coding and debugging with Lua ? I am new into Lua and I am using ZeroBrane Studio, but not sure if it is the best option.

griegler / octnetfusion

How to run train_s1n02_tsdfhist? #2