wangshaosun / grasp-transformer Goto Github PK

Can you provide some examples of multi-object. Is it generated from dataset or real scene.

Hi,
Is it possible to get the pre-trained weights for different datasets?

Hello, can you provide your trained model. I just want to know the parameters of your model.

Hello, it is very nice work, and inspired my work.
I want to apply this model to our panda robot, but I could't find the files.
So, could you share your panda simulation files? It will be exactly help me a lot.

您好，我是南京信息工程大学的一名在读硕士，最近再复现您发表在IEEE ROBOTICS AND AUTOMATION LETTERS上的论文When Transformer Meets Robotic Grasping: Exploits Context for Efficient Grasp Detection，有关TF-Grasp的代码中有一点疑问想向您请教育一下：

我在evaluation.py的代码中只找到了IoU>0.25这个条件，没有找到预测抓取框与gt抓取框角度相差<30度这个条件，在GGCNN的源码中似乎也没有找到这部分内容，所以来向您请教一下，万分感谢！

能否提供一下checkpoints，谢谢！

请问老哥，最后的论文报点精度是如何得出的呢，是如何评估出来的，还是说就是在训练过程中的验证最高精度呢

Hello, Mr. Wang. Thank you for your code sharing.
I have a questions that confuse me:
when I run main.py, it have error " 'Namespace‘ object has no attribute ‘vis‘ ".
So I checked “ gg_cnn”，find the main.py don't have " parser.add_argument('--vis', action='store_true', help='Visualise the training process') ".
After I added it, I found it was working normally.But when running in this way, the output models are always “ epoch_XX_iou_0 ”, and every epoches runs very fast.
I wonder if some of my parameters are not set, resulting in no training.
I hope you can take the time to answer my confusion. Thanlk you very much!!!!
In addition, I noticed that you are Chinese. This is my wechat17638091261.
I really hope we can have further communication,
Finally, praise your scientific research spirit

师兄，有没有你在grasp 1b上的结果？那个评估是按照最高得分抓取框评价，还是说top number 评价，还是说得分高于多少的全部评价？

Hello,Mr.Wang
I would like to know after the model is generated, is there a test or visualization phase for this project?
Looking forward to your answer.

Hello. Thank you for sharing. It is very helpful to me. I have a question about your paper.
On the Cornell grasp dataset, how does your code reflect the accuracy of image-wise and object-wise? In other words, in your code, where can it tell me the type of accuracy？ Object or image?
I'm a beginner. I have also checked the code of other projects, but I still haven't found the answer to the question。
Hope to get your answer！！！

I keep getting this error that it could not find the function center but I see it being there in the under the GraspRectangle class
when I try to train the Cornell grasp dataset using the main.py

AttributeError: Caught AttributeError in DataLoader worker process 0.
Original Traceback (most recent call last):
  File "/data/ali/grasp-transformer/env/lib/python3.8/site-packages/torch/utils/data/_utils/worker.py", line 302, in _worker_loop
    data = fetcher.fetch(index)
  File "/data/ali/grasp-transformer/env/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 58, in fetch
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/data/ali/grasp-transformer/env/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 58, in <listcomp>
    data = [self.dataset[idx] for idx in possibly_batched_index]
  File "/data/ali/grasp-transformer/utils/data/grasp_data.py", line 69, in __getitem__
    rgb_img = self.get_rgb(idx, rot, zoom_factor)
  File "/data/ali/grasp-transformer/utils/data/cornell_data.py", line 69, in get_rgb
    center, left, top = self._get_crop_attrs(idx)
  File "/data/ali/grasp-transformer/utils/data/cornell_data.py", line 43, in _get_crop_attrs
    center = gtbbs.center
  File "/data/ali/grasp-transformer/utils/dataset_processing/grasp.py", line 44, in __getattr__
    raise AttributeError("Couldn't find function %s in BoundingBoxes or BoundingBox" % attr)
AttributeError: Couldn't find function center in BoundingBoxes or BoundingBox

@WangShaoSUN

Hi, dear Shaochen:

I cannot open the link when to reproduce the experiment on a robot provided by your link.