Enroot installation - training part not working

[MAli7319]

Hi. I recently tried to run your project by utilizing enroot since I do not have ubuntu 20.

After running this command:
./devel/lib/neural_mapping/neural_mapping_node train src/M2Mapping/config/replica/replica.yaml src/M2Mapping/data/Replica/room2

I encountered this error message:

TORCH_VERSION: 2.4.1
Using path derived from config file: "src/M2Mapping"
output_path: "src/M2Mapping/output/2025-03-29-20-29-32_room2_replica.yaml"
data_path: "src/M2Mapping/data/Replica/room2"
print_files: src/M2Mapping/output/2025-03-29-20-29-32_room2_replica.yaml/config/scene/replica.yaml
%YAML:1.0

rosrun neural_mapping neural_mapping_node src/RIM2/neural_mapping/config/neural_rgbd/neural_rgbd.yaml src/RIM2/data/neural_rgbd_data/kitchen_kitti_format

base_config: "../base.yaml"

iter_step: 20000

trace_iter: 50 # sphere tracing iteration times

preload: 1 # 0: disable; 1: enable # accelerate the loading process but will cost more memory
llff: 0 # 0: disable; 1: enable; every 8 frame will be used for evaluation
cull_mesh: 1
prob_map_en: 0 # 0: disable; 1: enable; Whether the map is a probabilistic map

dataset_type

Replica = 0,

R3live = 1,

NeuralRGBD = 2,

KITTI = 3,

FastLivo = 4,

dataset_type: 0

dir_embedding_degree: 0

map:
map_origin: !!opencv-matrix
rows: 1
cols: 3
dt: d
data: [ 0, 0, 0 ]
map_size: 14

min_range: 0.0
max_range: 100
ds_pt_num: 10000 # downsampled point number
max_pt_num: 1000000

leaf_sizes: 0.05
fill_level: 1
bce_sigma: 0.02
sphere_trace_thr: 1e-3

outlier_removal_interval: 2000
outlier_remove: 0 # unnecessary for static scenes
outlier_dist: 0.05

visualization

vis_attribute: 2 # 0: disable to save storage; 1: normal; 2: color
vis_resolution: 0.1 # better no more than leaf_sizes or will miss faces
export_resolution: 0.01
fps: 30
print_files end
print_files: src/M2Mapping/output/2025-03-29-20-29-32_room2_replica.yaml/config/base.yaml
%YAML:1.0

debug: 0 # 0: disable; 1: enable

device_param: 1 # 0: cpu; 1: gpu

tcnn encoder params

n_levels: 16
n_features_per_level: 2
log2_hashmap_size: 19

tcnn decoder params

hidden_dim: 64
geo_feat_dim: 14 # geo_feat_dim + k_strc_dim <= 16 / 8 / 4 / 2 or tcnn decoder will become cutlass and crash
geo_num_layer: 3
color_hidden_dim: 64
color_num_layer: 3

rim params

trunc_sdf: 1
surface_sample_num: 3
free_sample_num: 3
color_batch_pt_num: 256000 # color render pt batch size

lr: 5e-3
lr_end: 1e-4

sdf_batch_ray_num: 2048
color_batch_pt_num: 25600
sdf_weight: 1.0
eikonal_weight: 1e-1 # it will greatly affect structure
curvate_weight: 5e-4 # should be the same loss level to eikonal loss
dist_weight: 1e-1

rgb_weight: 10.0

visualization

vis_frame_step: 10

export_interval: 1000 # every export_interval frames, the test will be exported
export_colmap_format: 0 # 0: disable; 1: for 3dgs; 2: for nerfstudio colmap
export_train_pcl: 0 # 0: disable; 1: enable
print_files end
print_files: src/M2Mapping/config/replica/replica.yaml
%YAML:1.0

rosrun neural_mapping neural_mapping_node src/RIM2/neural_mapping/config/neural_rgbd/neural_rgbd.yaml src/RIM2/data/neural_rgbd_data/kitchen_kitti_format

base_config: "../base.yaml"

iter_step: 20000

trace_iter: 50 # sphere tracing iteration times

preload: 1 # 0: disable; 1: enable # accelerate the loading process but will cost more memory
llff: 0 # 0: disable; 1: enable; every 8 frame will be used for evaluation
cull_mesh: 1
prob_map_en: 0 # 0: disable; 1: enable; Whether the map is a probabilistic map

dataset_type

Replica = 0,

R3live = 1,

NeuralRGBD = 2,

KITTI = 3,

FastLivo = 4,

dataset_type: 0

dir_embedding_degree: 0

map:
map_origin: !!opencv-matrix
rows: 1
cols: 3
dt: d
data: [ 0, 0, 0 ]
map_size: 14

min_range: 0.0
max_range: 100
ds_pt_num: 10000 # downsampled point number
max_pt_num: 1000000

leaf_sizes: 0.05
fill_level: 1
bce_sigma: 0.02
sphere_trace_thr: 1e-3

outlier_removal_interval: 2000
outlier_remove: 0 # unnecessary for static scenes
outlier_dist: 0.05

visualization

vis_attribute: 2 # 0: disable to save storage; 1: normal; 2: color
vis_resolution: 0.1 # better no more than leaf_sizes or will miss faces
export_resolution: 0.01
fps: 30
print_files end
Load 2000 data.
2000
Error: [enforce fail at alloc_cpu.cpp:117] err == 0. DefaultCPUAllocator: can't allocate memory: you tried to allocate 19584000000 bytes. Error code 12 (Cannot allocate memory)
frame #0: c10::ThrowEnforceNotMet(char const*, int, char const*, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, void const*) + 0x68 (0x7e57984e2618 in /root/libtorch/lib/libc10.so)
frame #1: c10::alloc_cpu(unsigned long) + 0x39f (0x7e57984d038f in /root/libtorch/lib/libc10.so)
frame #2: + 0x2e3f7 (0x7e579848f3f7 in /root/libtorch/lib/libc10.so)
frame #3: + 0x191c9dc (0x7e5780f1c9dc in /root/libtorch/lib/libtorch_cpu.so)
frame #4: at::detail::empty_generic(c10::ArrayRef, c10::Allocator*, c10::DispatchKeySet, c10::ScalarType, std::optionalc10::MemoryFormat) + 0x28 (0x7e5780f16598 in /root/libtorch/lib/libtorch_cpu.so)
frame #5: at::detail::empty_cpu(c10::ArrayRef, c10::ScalarType, bool, std::optionalc10::MemoryFormat) + 0x57 (0x7e5780f16617 in /root/libtorch/lib/libtorch_cpu.so)
frame #6: at::detail::empty_cpu(c10::ArrayRef, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional, std::optionalc10::MemoryFormat) + 0x49 (0x7e5780f16689 in /root/libtorch/lib/libtorch_cpu.so)
frame #7: at::native::empty_cpu(c10::ArrayRef, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional, std::optionalc10::MemoryFormat) + 0x34 (0x7e578173d404 in /root/libtorch/lib/libtorch_cpu.so)
frame #8: + 0x2eafc00 (0x7e57824afc00 in /root/libtorch/lib/libtorch_cpu.so)
frame #9: + 0x2eafc7f (0x7e57824afc7f in /root/libtorch/lib/libtorch_cpu.so)
frame #10: at::_ops::empty_memory_format::redispatch(c10::DispatchKeySet, c10::ArrayRefc10::SymInt, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional, std::optionalc10::MemoryFormat) + 0xfb (0x7e578205df7b in /root/libtorch/lib/libtorch_cpu.so)
frame #11: + 0x2e73283 (0x7e5782473283 in /root/libtorch/lib/libtorch_cpu.so)
frame #12: at::_ops::empty_memory_format::call(c10::ArrayRefc10::SymInt, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional, std::optionalc10::MemoryFormat) + 0x1b3 (0x7e57820ad383 in /root/libtorch/lib/libtorch_cpu.so)
frame #13: at::native::zeros_symint(c10::ArrayRefc10::SymInt, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional) + 0x17d (0x7e57817458bd in /root/libtorch/lib/libtorch_cpu.so)
frame #14: + 0x3083ea9 (0x7e5782683ea9 in /root/libtorch/lib/libtorch_cpu.so)
frame #15: at::_ops::zeros::redispatch(c10::DispatchKeySet, c10::ArrayRefc10::SymInt, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional) + 0xf6 (0x7e5781cb5936 in /root/libtorch/lib/libtorch_cpu.so)
frame #16: + 0x2e70cc9 (0x7e5782470cc9 in /root/libtorch/lib/libtorch_cpu.so)
frame #17: at::_ops::zeros::call(c10::ArrayRefc10::SymInt, std::optionalc10::ScalarType, std::optionalc10::Layout, std::optionalc10::Device, std::optional) + 0x1b2 (0x7e5781d16f32 in /root/libtorch/lib/libtorch_cpu.so)
frame #18: torch::zeros(c10::ArrayRef, c10::TensorOptions) + 0x1ad (0x7e579862795d in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #19: dataparser::DataParser::load_colors(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, bool, bool const&) + 0x3d3 (0x7e571d2e68b3 in /root/m2mapping_ws/devel/lib/libdata_parser.so)
frame #20: dataparser::Replica::load_data() + 0x17f (0x7e577f5e431f in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #21: dataparser::Replica::Replica(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, c10::Device const&, bool const&, float const&) + 0x83c (0x7e577f5e551c in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #22: dataloader::DataLoader::DataLoader(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, int const&, c10::Device const&, bool const&, float const&, sensor::Sensors const&) + 0x638 (0x7e577f5d4248 in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #23: NeuralSLAM::NeuralSLAM(int const&, std::filesystem::__cxx11::path const&, std::filesystem::__cxx11::path const&) + 0x24c (0x7e579861005c in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #24: main + 0x680 (0x5d41ac6a5eb0 in ./devel/lib/neural_mapping/neural_mapping_node)
frame #25: __libc_start_main + 0xf3 (0x7e571d450083 in /lib/x86_64-linux-gnu/libc.so.6)
frame #26: _start + 0x2e (0x5d41ac6a622e in ./devel/lib/neural_mapping/neural_mapping_node)

double free or corruption (fasttop)
Segmentation fault (core dumped)

What is the solution? Thanks in advance.

[jianhengLiu]

It is likely out of RAM. Try to set 'preload: 0' to disable preloading training data into RAM in https://github.com/hku-mars/M2Mapping/blob/main/config/replica/replica.yaml#L10

[MAli7319]

This error is solved but another one is occured.

Load 2000 data.
2000
Load 2000 data.
[2000, 10000, 1]
Error: Pose file does not exist: src/M2Mapping/data/Replica/room2/eval/traj.txt
double free or corruption (fasttop)
Segmentation fault (core dumped)

traj.txt is directly under room2, not eval. How can I change the path?

[MAli7319]

I ve tried to make eval directory and move the traj.txt into it. But it raised this error:

visualization

vis_attribute: 2 # 0: disable to save storage; 1: normal; 2: color
vis_resolution: 0.1 # better no more than leaf_sizes or will miss faces
export_resolution: 0.01
fps: 30
print_files end
pose_path_ does not exist: "src/M2Mapping/data/Replica/room2/traj.txt"
Training mode initialized with config: src/M2Mapping/config/replica/replica.yaml
Starting occupancy map building from raw data...
terminate called after throwing an instance of 'c10::Error'
what(): Expected a proper Tensor but got None (or an undefined Tensor in C++) for argument #0 'self'
Exception raised from checked_cast_variable at ../torch/csrc/autograd/VariableTypeManual.cpp:66 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits, std::allocator >) + 0xb0 (0x736422f9d7f0 in /root/libtorch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&) + 0xfa (0x736422f45f7e in /root/libtorch/lib/libc10.so)
frame #2: + 0x61f8829 (0x73648aff8829 in /root/libtorch/lib/libtorch_cpu.so)
frame #3: + 0x4e8a4cb (0x736489c8a4cb in /root/libtorch/lib/libtorch_cpu.so)
frame #4: + 0x4e8abbe (0x736489c8abbe in /root/libtorch/lib/libtorch_cpu.so)
frame #5: at::_ops::view::call(at::Tensor const&, c10::ArrayRefc10::SymInt) + 0x183 (0x736487a7d773 in /root/libtorch/lib/libtorch_cpu.so)
frame #6: NeuralSLAM::build_occ_map() + 0xd1 (0x73649dce74e1 in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #7: NeuralSLAM::batch_train() + 0xe7 (0x73649dcedda7 in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #8: + 0xd6df4 (0x736422e10df4 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #9: + 0x8609 (0x73649dbfa609 in /lib/x86_64-linux-gnu/libpthread.so.0)
frame #10: clone + 0x43 (0x736422c67133 in /lib/x86_64-linux-gnu/libc.so.6)

Stack trace (most recent call last) in thread 49105:
#16 Object "", at 0xffffffffffffffff, in
#15 Object "/usr/lib/x86_64-linux-gnu/libc-2.31.so", at 0x736422c67132, in clone
#14 Object "/usr/lib/x86_64-linux-gnu/libpthread-2.31.so", at 0x73649dbfa608, in start_thread
#13 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28", at 0x736422e10df3, in
#12 Source "/root/m2mapping_ws/src/M2Mapping/include/neural_mapping/neural_mapping.cpp", line 876, in batch_train [0x73649dcedda6]
873: }
874: static auto p_timer = llog::CreateTimer("build_occ_map");
875: p_timer->tic();
> 876: build_occ_map();
877:
878: torch::optim::AdamOptions adam_options;
879: adam_options.lr(k_lr);
#11 | Source "/root/m2mapping_ws/src/M2Mapping/include/neural_mapping/neural_mapping.cpp", line 529, in view
| 528: auto pcl_depth =
| > 529: data_loader_ptr->dataparser_ptr->train_depth_pack.depth.view({-1});
| 530: auto valid_mask = (pcl_depth > k_min_range) & (pcl_depth < k_max_range);
| 531: auto valid_idx = valid_mask.nonzero().squeeze();
Source "/root/libtorch/include/ATen/core/TensorBody.h", line 4349, in build_occ_map [0x73649dce74e0]
4347: // aten::view(Tensor(a) self, SymInt[] size) -> Tensor(a)
4348: inline at::Tensor Tensor::view(at::IntArrayRef size) const {
>4349: return at::ops::view::call(const_cast<Tensor&>(*this), c10::fromIntArrayRefSlow(size));
4350: }
4351:
4352: // aten::view(Tensor(a) self, SymInt[] size) -> Tensor(a)
#10 Object "/root/libtorch/lib/libtorch_cpu.so", at 0x736487a7d772, in at::ops::view::call(at::Tensor const&, c10::ArrayRefc10::SymInt)
#9 Object "/root/libtorch/lib/libtorch_cpu.so", at 0x736489c8abbd, in c10::impl::wrap_kernel_functor_unboxed<c10::impl::detail::WrapFunctionIntoFunctor<c10::CompileTimeFunctionPointer<at::Tensor (c10::DispatchKeySet, at::Tensor const&, c10::ArrayRefc10::SymInt), &torch::autograd::VariableType::(anonymous namespace)::view>, at::Tensor, c10::guts::typelist::typelist<c10::DispatchKeySet, at::Tensor const&, c10::ArrayRefc10::SymInt > >, at::Tensor (c10::DispatchKeySet, at::Tensor const&, c10::ArrayRefc10::SymInt)>::call(c10::OperatorKernel*, c10::DispatchKeySet, at::Tensor const&, c10::ArrayRefc10::SymInt)
#8 Object "/root/libtorch/lib/libtorch_cpu.so", at 0x736489c8a4ca, in torch::autograd::VariableType::(anonymous namespace)::view(c10::DispatchKeySet, at::Tensor const&, c10::ArrayRefc10::SymInt)
#7 Object "/root/libtorch/lib/libtorch_cpu.so", at 0x73648aff8828, in torch::autograd::VariableType::(anonymous namespace)::checked_cast_variable(at::Tensor const&, char const*, int)
#6 Object "/root/libtorch/lib/libc10.so", at 0x736422f45fa5, in c10::detail::torchCheckFail(char const*, char const*, unsigned int, std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&)
#5 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28", at 0x736422de4698, in __cxa_throw
#4 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28", at 0x736422de43e6, in std::terminate()
#3 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28", at 0x736422de437b, in
#2 Object "/usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28", at 0x736422dd88d0, in
#1 Object "/usr/lib/x86_64-linux-gnu/libc-2.31.so", at 0x736422b6a858, in abort
#0 Object "/usr/lib/x86_64-linux-gnu/libc-2.31.so", at 0x736422b8b00b, in raise
Aborted (Signal sent by tkill() 49080 0)
Aborted (core dumped)

[MAli7319]

Instead of moving the traj.txt, I copied it into the eval directory and also there is a need for results folder under the eval directory. It solved the problem but another one occured again.

TORCH_VERSION: 2.4.1
Using path derived from config file: "src/M2Mapping"
output_path: "src/M2Mapping/output/2025-03-30-00-33-19_room2_replica.yaml"
data_path: "src/M2Mapping/data/Replica/room2"
print_files: src/M2Mapping/output/2025-03-30-00-33-19_room2_replica.yaml/config/scene/replica.yaml
%YAML:1.0

rosrun neural_mapping neural_mapping_node src/RIM2/neural_mapping/config/neural_rgbd/neural_rgbd.yaml src/RIM2/data/neural_rgbd_data/kitchen_kitti_format

base_config: "../base.yaml"

iter_step: 20000

trace_iter: 50 # sphere tracing iteration times

preload: 0 # 0: disable; 1: enable # accelerate the loading process but will cost more memory
llff: 0 # 0: disable; 1: enable; every 8 frame will be used for evaluation
cull_mesh: 1
prob_map_en: 0 # 0: disable; 1: enable; Whether the map is a probabilistic map

dataset_type

Replica = 0,

R3live = 1,

NeuralRGBD = 2,

KITTI = 3,

FastLivo = 4,

dataset_type: 0

dir_embedding_degree: 0

map:
map_origin: !!opencv-matrix
rows: 1
cols: 3
dt: d
data: [ 0, 0, 0 ]
map_size: 14

min_range: 0.0
max_range: 100
ds_pt_num: 10000 # downsampled point number
max_pt_num: 1000000

leaf_sizes: 0.05
fill_level: 1
bce_sigma: 0.02
sphere_trace_thr: 1e-3

outlier_removal_interval: 2000
outlier_remove: 0 # unnecessary for static scenes
outlier_dist: 0.05

visualization

vis_attribute: 2 # 0: disable to save storage; 1: normal; 2: color
vis_resolution: 0.1 # better no more than leaf_sizes or will miss faces
export_resolution: 0.01
fps: 30
print_files end
print_files: src/M2Mapping/output/2025-03-30-00-33-19_room2_replica.yaml/config/base.yaml
%YAML:1.0

debug: 0 # 0: disable; 1: enable

device_param: 1 # 0: cpu; 1: gpu

tcnn encoder params

n_levels: 16
n_features_per_level: 2
log2_hashmap_size: 19

tcnn decoder params

hidden_dim: 64
geo_feat_dim: 14 # geo_feat_dim + k_strc_dim <= 16 / 8 / 4 / 2 or tcnn decoder will become cutlass and crash
geo_num_layer: 3
color_hidden_dim: 64
color_num_layer: 3

rim params

trunc_sdf: 1
surface_sample_num: 3
free_sample_num: 3
color_batch_pt_num: 256000 # color render pt batch size

lr: 5e-3
lr_end: 1e-4

sdf_batch_ray_num: 2048
color_batch_pt_num: 25600
sdf_weight: 1.0
eikonal_weight: 1e-1 # it will greatly affect structure
curvate_weight: 5e-4 # should be the same loss level to eikonal loss
dist_weight: 1e-1

rgb_weight: 10.0

visualization

vis_frame_step: 10

export_interval: 1000 # every export_interval frames, the test will be exported
export_colmap_format: 0 # 0: disable; 1: for 3dgs; 2: for nerfstudio colmap
export_train_pcl: 0 # 0: disable; 1: enable
print_files end
print_files: src/M2Mapping/config/replica/replica.yaml
%YAML:1.0

rosrun neural_mapping neural_mapping_node src/RIM2/neural_mapping/config/neural_rgbd/neural_rgbd.yaml src/RIM2/data/neural_rgbd_data/kitchen_kitti_format

base_config: "../base.yaml"

iter_step: 20000

trace_iter: 50 # sphere tracing iteration times

preload: 0 # 0: disable; 1: enable # accelerate the loading process but will cost more memory
llff: 0 # 0: disable; 1: enable; every 8 frame will be used for evaluation
cull_mesh: 1
prob_map_en: 0 # 0: disable; 1: enable; Whether the map is a probabilistic map

dataset_type

Replica = 0,

R3live = 1,

NeuralRGBD = 2,

KITTI = 3,

FastLivo = 4,

dataset_type: 0

dir_embedding_degree: 0

map:
map_origin: !!opencv-matrix
rows: 1
cols: 3
dt: d
data: [ 0, 0, 0 ]
map_size: 14

min_range: 0.0
max_range: 100
ds_pt_num: 10000 # downsampled point number
max_pt_num: 1000000

leaf_sizes: 0.05
fill_level: 1
bce_sigma: 0.02
sphere_trace_thr: 1e-3

outlier_removal_interval: 2000
outlier_remove: 0 # unnecessary for static scenes
outlier_dist: 0.05

visualization

vis_attribute: 2 # 0: disable to save storage; 1: normal; 2: color
vis_resolution: 0.1 # better no more than leaf_sizes or will miss faces
export_resolution: 0.01
fps: 30
print_files end
Load 2000 data.
2000
Load 2000 data.
[2000, 10000, 1]
Load 0 data.
Error: Expected eval_color_poses_.size(0) == eval_color_filelists_.size() to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.)
Exception raised from load_eval_data at /root/m2mapping_ws/src/M2Mapping/include/data_loader/data_parsers/replica_parser.hpp:59 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits, std::allocator >) + 0xb0 (0x70c2be19d7f0 in /root/libtorch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, char const*) + 0x84 (0x70c2be146092 in /root/libtorch/lib/libc10.so)
frame #2: dataparser::Replica::load_eval_data() + 0x32f (0x70c31ffc111f in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #3: dataparser::Replica::load_data() + 0x275 (0x70c31ffc1415 in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #4: dataparser::Replica::Replica(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, c10::Device const&, bool const&, float const&) + 0x83c (0x70c31ffc251c in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #5: dataloader::DataLoader::DataLoader(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, int const&, c10::Device const&, bool const&, float const&, sensor::Sensors const&) + 0x638 (0x70c31ffb1248 in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #6: NeuralSLAM::NeuralSLAM(int const&, std::filesystem::__cxx11::path const&, std::filesystem::__cxx11::path const&) + 0x24c (0x70c338ed805c in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #7: main + 0x680 (0x5a84931ceeb0 in /root/m2mapping_ws/devel/lib/neural_mapping/neural_mapping_node)
frame #8: __libc_start_main + 0xf3 (0x70c2bdd6c083 in /lib/x86_64-linux-gnu/libc.so.6)
frame #9: _start + 0x2e (0x5a84931cf22e in /root/m2mapping_ws/devel/lib/neural_mapping/neural_mapping_node)

double free or corruption (fasttop)
Segmentation fault (core dumped)

[jianhengLiu]

Sorry for the unclear instruction. Here is a data structure example of Replica Room-2, check if it is correct:

├── room2
│   ├── eval
│   │   └── results
│   │   │   └── *.jpg
│   │   │   └── *.png
│   │   └── traj.txt
│   └── results
│   │   └── *.jpg
│   │   └── *.png
│   └── traj.txt

[MAli7319]

room2/eval/results is empty. Other than that it is ok

[MAli7319]

The problem still exists. I just responded the data structure and it is not complete. Can you reopen this issue or at least, propose a solution?

[jianhengLiu]

Sorry, I thought it was work. Try to download the Replica evaluation data in https://furtive-lamprey-00b.notion.site/M2Mapping-Datasets-e6318dcd710e4a9d8a4f4b3fbe176764 to see if it will solve the case.

[MAli7319]

I noticed that commands in Data Preparation part 3.1 in README do not download the all data that had been said in the link you shared. But unzip commands for those are exist. I do not know whether or not it is intentional.

I downloaded

• and unzipped the replica_extra_eval and placed into /m2mapping_ws/src/M2Mapping/data
• and cull_replica_mesh into /m2mapping_ws/src/M2Mapping/data/Replica

as you implied with unzip command in 3.1

The error message is still the same.

Error: Expected eval_color_poses_.size(0) == eval_color_filelists_.size() to be true, but got false. (Could this error message be improved? If so, please report an enhancement request to PyTorch.)
Exception raised from load_eval_data at /root/m2mapping_ws/src/M2Mapping/include/data_loader/data_parsers/replica_parser.hpp:59 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits, std::allocator >) + 0xb0 (0x701af799d7f0 in /root/libtorch/lib/libc10.so)
frame #1: c10::detail::torchCheckFail(char const*, char const*, unsigned int, char const*) + 0x84 (0x701af7946092 in /root/libtorch/lib/libc10.so)
frame #2: dataparser::Replica::load_eval_data() + 0x32f (0x701b597e411f in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #3: dataparser::Replica::load_data() + 0x275 (0x701b597e4415 in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #4: dataparser::Replica::Replica(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, c10::Device const&, bool const&, float const&) + 0x83c (0x701b597e551c in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #5: dataloader::DataLoader::DataLoader(std::__cxx11::basic_string<char, std::char_traits, std::allocator > const&, int const&, c10::Device const&, bool const&, float const&, sensor::Sensors const&) + 0x638 (0x701b597d4248 in /root/m2mapping_ws/devel/lib/libdata_loader.so)
frame #6: NeuralSLAM::NeuralSLAM(int const&, std::filesystem::__cxx11::path const&, std::filesystem::__cxx11::path const&) + 0x24c (0x701b7271d05c in /root/m2mapping_ws/devel/lib/libneural_mapping_lib.so)
frame #7: main + 0x680 (0x61f42a19ceb0 in /root/m2mapping_ws/devel/lib/neural_mapping/neural_mapping_node)
frame #8: __libc_start_main + 0xf3 (0x701af756c083 in /usr/lib/x86_64-linux-gnu/libc.so.6)
frame #9: _start + 0x2e (0x61f42a19d22e in /root/m2mapping_ws/devel/lib/neural_mapping/neural_mapping_node)

double free or corruption (fasttop)
Segmentation fault (core dumped)

What should I do now?

[jianhengLiu]

Thanks for the suggestion, I will update the README later.
As shown in the error message "Expected eval_color_poses_.size(0) == eval_color_filelists_.size() to be true, but got false.", it is likely that the color images file number does not match the color poses number (in 'data/Replica/room2/eval/results’ and 'data/Replica/room2/eval/traj.txt'.
In your case running the Replica Room-2 scene, it is supposed with 184 file items under 'data/Replica/room2/eval/results’ and 92 lines in 'data/Replica/room2/eval/traj.txt'. Check if it matches the suggeted number.

[MAli7319]

There are 4000 images under 'data/Replica/room2/eval/results’ and 2000 lines in 'data/Replica/room2/eval/traj.txt' file. I thought traj.txt file under eval and under room2 are splitted equally but they are exactly the same. What do you suggest?

[jianhengLiu]

It seems that 'replica_extra_eval' is not in the right place.
First put 'Replica.zip', 'Replica cull_replica_mesh.zip' and 'replica_extra_eval.zip' under 'm2mapping_ws', then:

unzip -d src/M2Mapping/data Replica.zip
unzip -d src/M2Mapping/data/Replica cull_replica_mesh.zip
unzip -d src/M2Mapping/data replica_extra_eval.zip

[MAli7319]

Error: Pose file does not exist: src/M2Mapping/data/Replica/room2/eval/traj.txt
double free or corruption (fasttop)
Segmentation fault (core dumped)

I reinstalled and placed the data with the way you proposed. And it raised this error.
Further checking, replica_extra_eval file has room2/eval/traj.txt data path, not the Replica.zip itself. So in my opinion, the actual path should be like:
src/M2Mapping/data/replica_extra_eval/Replica/room2/eval/traj.txt

So should I cut the eval directory from src/M2Mapping/data/replica_extra_eval/Replica/room2/ and paste it to src/M2Mapping/data/Replica/room2/ or somehow, change the way the code looks for the traj.txt?