Sorry for bothering you. I run the file gt_single_object.py ,then create the folder data_linemod_gt, I run train_single.sh, then I got the following information:

Total BFLOPS 65.290
Loading weights from darknet53.conv.74...
seen 64
Done!
Learning Rate: 0.001, Momentum: 0.9, Decay: 0.0005
Couldn't open file: data_linemod_gt/02/all.txt

What should I do?

I am running everything on Windows and I've finally managed to get everything to run without errors. But I have the following problem when I run the betapose_evaluate.py, the method doesn't detect any objects in all images. I am almost sure the problem is in the first step (object detection with YOLO), because it is the one that detects the object.

Here is what I've done to create the yolo weights file:

1. I've compiled YOLO directly from its original source, because I am on windows and the make step that you've described wouldn't have worked for me. This means I didn't use the contents of the train_YOLO folder at all.
2. I've then trained the network for the APE object from the LINEMOD dataset. I've trained it with 2000 iterations
3. I've tested the weights with the YOLO program compiled in step 1) and it worked fine - it detected objects as expected.

BUT when I use the weight file in the betapose_evaluate.py it doesn't work. Can you tell me if you have made any major changes in the darknet code(train_YOLO)? Is my problem that I haven't used a darknet built from the train_YOLO folder? Here is a link to my weights for the APE object from the LINEMOD dataset - https://www.dropbox.com/s/2e3xdshsavp3t5a/01.weights?dl=0 and here is an example image with an APE in it - https://www.dropbox.com/s/rhar1p251dp4z31/000002.jpg?dl=0

P.S. Could you also recommend an open-sourced method for synthetic dataset generation?

Traceback (most recent call last):
  File "annotate_keypoint.py", line 360, in <module>
    bench = load_bench(sixd_base)
  File "annotate_keypoint.py", line 73, in load_bench
    bench.kpmodels['{:02d}'.format(ID)].load(os.path.join(base_path, 'kpmodels/' + name + '.ply'), scale=bench.scale_to_meters)
  File "/home/Workspace/Project/betapose/2_keypoint_annotator/utils/model.py", line 80, in load
    data = PlyData.read(path)
  File "/home/anaconda3/envs/pytorch/lib/python2.7/site-packages/plyfile.py", line 390, in read
    (must_close, stream) = _open_stream(stream, 'read')
  File "/home/anaconda3/envs/pytorch/lib/python2.7/site-packages/plyfile.py", line 469, in _open_stream
    return (True, open(stream, read_or_write[0] + 'b'))

What is kpmodels？You just say

Only folders called models and test are needed. Put them in DATAROOT/models and DATAROOT/test

Hi,
I would like to take an image, forward it after the training and show results on the original image.
what are the functions that apply the different transformations to the input image to make inference? and how can i return to the original image and project results on it?

As title, the dataset can not be downloaded.
Can you update the dataset download link?

thanks

Hi,

I tried to train your network on the Linemod Ape object (01). I followed every steps of your tutorial.
But when I run betapose_evaluate, I cannot find results as you mention in your paper, especially with ADD metric.
Do you have pkl (output of train KPD) file on the Ape object so I can check if the problem come from my learning ?

Regards

What is the license for this code? Lack of license implies full copyright protection.

How is the target detection tag get in the data set obtained?

https://github.com/sjtuytc/segmentation-driven-pose
It's far more interesting.

Hello, sjtuytc.

Thanks for you grate work and sharing.
I want to run your code(train and test), but as you said we need to prepare data by ourself.
I will try to prepare data as darknet , but can you show me the whole dataset folder structure? so that I can more efficiently do this work.

Thanks,
Yuning

Please tell me how to prepare data as AlexeyAB/darknet's instructions? I download the darknet'code, but the "README.md" in ./scripts folder just some website of dataset. I don't know how to prepare data as AlexeyAB/darknet's instructions.

Ⅳ. Training
3.Train on Linemod dataset without DPG.
when i try $ python src/train.py --trainBatch 28 --expID seq5_Nov_1_1 --optMethod adam
i get this error
FileNotFoundError: [Errno 2] No such file or directory: '../exp/coco/seq5_Nov_1_1'

please tell me how to solve it ,thank you!

"when on the Occlusion dataset we only keep l = 10 most confident keypoints and abandon other low-confident keypoints".
hi~ I wonder how you select the most 10 confident keypoint for pnp? what's the confidence based on ?
I can't find the relevant code, could you give me some help?
best wishes.

As title, when I tried to load sixd data I met some problems.

You load sixd data from 'gt.yml'(Is it from scene_gt.json?) In sixd.py, and this code: fr.gt.append((gt['obj_id'], pose, gt['obj_bb'])) for loading pose and ground truth bbox. But in gt.yml file I can't find 'obj_bb' items.
I also try to delete this item and just load pose, but It seems use for cropping image.
can you give me some suggestions?

thanks,
yuning

Hello I am trying to train a KPD, but I get this error:

Traceback (most recent call last): File "src/train.py", line 14, in <module> from evaluation import prediction File "...\betapose\3_6Dpose_estimator\train_KPD\src\evaluation.py", line 9, in <module> from predict.annot.coco_minival import Mscoco_minival ModuleNotFoundError: No module named 'predict.annot'

Could you please upload this missing file?

Scanning dependencies of target pcl-sift [ 50%] Building CXX object CMakeFiles/pcl-sift.dir/main.cpp.o [100%] Linking CXX executable pcl-sift /usr/bin/ld: warning: libboost_system.so.1.58.0, needed by /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../x86_64-linux-gnu/libpcl_common.so, may conflict with libboost_system.so.1.60.0 /usr/bin/ld: warning: libboost_thread.so.1.58.0, needed by /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../x86_64-linux-gnu/libpcl_common.so, may conflict with libboost_thread.so.1.60.0 /usr/bin/ld: warning: libboost_filesystem.so.1.58.0, needed by /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../x86_64-linux-gnu/libpcl_io.so, may conflict with libboost_filesystem.so.1.60.0 /usr/bin/ld: warning: libboost_iostreams.so.1.58.0, needed by /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../x86_64-linux-gnu/libpcl_io.so, may conflict with libboost_iostreams.so.1.60.0 /usr/bin/ld: warning: libboost_chrono.so.1.58.0, needed by /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../x86_64-linux-gnu/libpcl_io.so, may conflict with libboost_chrono.so.1.60.0 CMakeFiles/pcl-sift.dir/main.cpp.o：在函数‘main’中： /home/panxiang/ldd/betapose-master/1_keypoint_designator/main.cpp:93：对‘pcl::visualization::PCLVisualizer::PCLVisualizer(std::string const&, bool)’未定义的引用 /home/panxiang/ldd/betapose-master/1_keypoint_designator/main.cpp:101：对‘pcl::visualization::PCLVisualizer::setPointCloudRenderingProperties(int, double, std::string const&, int)’未定义的引用 CMakeFiles/pcl-sift.dir/main.cpp.o：在函数‘pcl::PLYWriter::write(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, bool)’中： /usr/include/pcl-1.7/pcl/io/ply_io.h:630：对‘pcl::PLYWriter::writeBinary(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, bool)’未定义的引用 /usr/include/pcl-1.7/pcl/io/ply_io.h:632：对‘pcl::PLYWriter::writeASCII(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, int, bool)’未定义的引用 CMakeFiles/pcl-sift.dir/main.cpp.o：在函数‘pcl::PLYWriter::write(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, bool, bool)’中： /usr/include/pcl-1.7/pcl/io/ply_io.h:653：对‘pcl::PLYWriter::writeBinary(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, bool)’未定义的引用 /usr/include/pcl-1.7/pcl/io/ply_io.h:655：对‘pcl::PLYWriter::writeASCII(std::string const&, pcl::PCLPointCloud2 const&, Eigen::Matrix<float, 4, 1, 0, 4, 1> const&, Eigen::Quaternion<float, 0> const&, int, bool)’未定义的引用 CMakeFiles/pcl-sift.dir/main.cpp.o：在函数‘pcl::search::OrganizedNeighbor<pcl::PointXYZ>::OrganizedNeighbor(bool, float, unsigned int)’中： /usr/include/pcl-1.7/pcl/search/organized.h:95：对‘pcl::search::Search<pcl::PointXYZ>::Search(std::string const&, bool)’未定义的引用 collect2: error: ld returned 1 exit status CMakeFiles/pcl-sift.dir/build.make:352: recipe for target 'pcl-sift' failed make[2]: *** [pcl-sift] Error 1 CMakeFiles/Makefile2:72: recipe for target 'CMakeFiles/pcl-sift.dir/all' failed make[1]: *** [CMakeFiles/pcl-sift.dir/all] Error 2 Makefile:83: recipe for target 'all' failed make: *** [all] Error 2 build_and_run.sh: 5: build_and_run.sh: ./pcl-sift: not found **I have encountered this problem, how can I solve it?**

Hello, I want to run your code. I met this problem.Can you tell me why?

Sorry for disturbance. Can you provide the file structure in './betapose/3_6Dpose_estimator/train_YOLO/scripts'?
In AlexeyAB/darknet the file structure is
├── VOCdevkit
│ └── VOC2007
│ ├── Annotations
│ │ ├──.xml
│ │ └── .xml
│ ├── ImageSets
│ │ ├── Layout
│ │ ├── Main
│ │ │ ├── test.txt
│ │ │ ├── train.txt
│ │ │ └── val.txt
│ │ └── Segmentation
│ ├── JPEGImages
│ │ ├──.jpg
│ │ └──.jpg
│ └── labels
│ └── .txt
└── voc_label.py

but you hadn't provide the voc_label.py and .xmlfiles

Thank you for your help. I have new problems again.
Create new model ############# Starting Epoch 0 ############# 0%| | 0/17 [00:00<?, ?it/s]Traceback (most recent call last): File "src/train.py", line 226, in <module> main() File "src/train.py", line 171, in main loss, acc = train(train_loader, m, criterion, optimizer, writer) File "src/train.py", line 27, in train for i, (inps, labels, setMask, imgset) in enumerate(train_loader_desc): File "/home/ltr/anaconda3/envs/betapose/lib/python3.6/site-packages/tqdm/_tqdm.py", line 1022, in __iter__ for obj in iterable: File "/home/ltr/anaconda3/envs/betapose/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 286, in __next__ return self._process_next_batch(batch) File "/home/ltr/anaconda3/envs/betapose/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 307, in _process_next_batch raise batch.exc_type(batch.exc_msg) RuntimeError: Traceback (most recent call last): File "/home/ltr/anaconda3/envs/betapose/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 57, in _worker_loop samples = collate_fn([dataset[i] for i in batch_indices]) File "/home/ltr/anaconda3/envs/betapose/lib/python3.6/site-packages/torch/utils/data/dataloader.py", line 57, in <listcomp> samples = collate_fn([dataset[i] for i in batch_indices]) File "/media/ltr/新加卷2/betapose/3_6Dpose_estimator/train_KPD/src/utils/dataset/coco.py", line 75, in __getitem__ 'coco', sf, self, train=self.is_train) File "/media/ltr/新加卷2/betapose/3_6Dpose_estimator/train_KPD/src/utils/pose.py", line 142, in generateSampleBox inp = rotate(inp) File "/media/ltr/新加卷2/betapose/3_6Dpose_estimator/train_KPD/src/torchsample/torchsample/transforms/affine_transforms.py", line 322, in __call__ center=True) File "/media/ltr/新加卷2/betapose/3_6Dpose_estimator/train_KPD/src/torchsample/torchsample/utils.py", line 131, in th_affine2d x_transformed = th_bilinear_interp2d(x.contiguous(), new_coords) File "/media/ltr/新加卷2/betapose/3_6Dpose_estimator/train_KPD/src/torchsample/torchsample/utils.py", line 167, in th_bilinear_interp2d x0_ix = x0.mul(stride[1]).long() RuntimeError: Expected object of type torch.FloatTensor but found type torch.LongTensor for argument #2 'other'