dc1991 / g2l_net Goto Github PK
View Code? Open in Web Editor NEWLicense: MIT License
License: MIT License
What should I be expecting from the demo output?
I tried running the demo, but I got the following error: AttributeError: 'NoneType' object has no attribute 'shape'
I then created a new valseg.lst with numbers 0~19 (since there are only 20 demo images); the demo code worked but there was no output.
Hello author, may I ask how to test the fps of the program?
Should it be tested on ROS or just run on the dataset?
Can you tell me the general method? I'm doing a comparative test.
Thank you!
can't import function from neighbor folder. Can you please explain the folder structure?
Hi, @DC1991 !Thanks for your excellent work!
Since the depth image in the Linemod seems too dark and couln't see anything, I wonder how to visualize the depth image in Figure 1. shown as follows .
Do you pre-process the original depth image by ajusting brightness or contrast?
Hi, thank you for your great work. Will you provide the completed training dataset? Or how do I generate it myself?
Thank you very much.
你好,我想问一下对于不同的物体,是需要训练不同的模型。还是说多个物体可以共用一个网络结构呢?
Thank you for your excellent work and sharing!
I'm trying to run the code and I have some questions, Would you mind give me some explaination ?
how to understand 'cors' in the code: output0, cors = model(imgs.cuda(), test=1)
, ? I find that the 'cors ' are also used in the following :
DC = int(W * cors[2][1][io] / 52)
DR = int(H * cors[2][0][io] / 52)
and how to undertand the 300 and 102 in the following code ?
dep3d = dep3d[np.where(dep3d[:, 2] > 300.0)]
dep3d = chooselimt_test(dep3d, 102, cen_depth)
Hope for your reply!
你好,我根据你提供的物体6的训练数据集,对模型进行了训练,发现网络对旋转的预测效果不太理想。我应该怎么做才能提高准确率呢?非常感谢你的回复。
Hello @DC1991
Thank you for your work.
I run your code and it executed well.
Unfortunately, I don't understand the meaning of output values (R and T). Would you mind give me some explaination ?
In your paper I found this part. Is the code to manipulate including in current source code?
If it isn't included, would you mind give me a link to the code?
However, both LINEMOD and YCB-Video datasets do
not contain the label for each point of the point cloud. To
train G2L-Net in a supervised fashion, we adopt an automatic way to label each point of the point cloud of [?]. As
described in [?], we label each point in two steps§ First, for
the 3D model of an object, we transform it into the camera
coordinate using the corresponding ground truth. We adopt
the implementation provided by [14] for this process.
Thank you
In train_G2L.py line 12
from G2L_Net.utils.networks_arch import *
I wanna know where is the G2L_Net?Can u give me a hand ?best wishes
我改好几个路径错误之后,遇到一个很怪异的问题:
`Traceback (most recent call last):
File "/home/leon/code project test/G2L_Net/demo/test_linemod.py", line 289, in
R, T = test(rgbs, deps, idx, model, classifier, classifier_ce, classifier_box, classifier_box_gan, classifier_box_vec, opt, pc, OR, Rt, Tt, imgid=idxx, temp=temp)
File "/home/leon/code project test/G2L_Net/demo/test_linemod.py", line 152, in test
imgs = letterbox(rgb[0], [416, 416], color=(127.5, 127.5, 127.5))
File "/home/leon/code project test/G2L_Net/demo/test_linemod.py", line 78, in letterbox
shape = img.shape[:2] # shape = [height, width]
AttributeError: 'NoneType' object has no attribute 'shape'`
我查看了"../test_sequence/01/rgb" 确认数据存在和路径没有问题,但是我无法理解为何依旧存在这样的错误。
而且我在运行test_linemod.py需要等待好几分钟才会出现报错的结果。
Hi, @DC1991 . Thanks for your outstanding work and sharing! I have trained the 'cat'(models 6 in this project), but I don't know how to make best_6.pt by myself. Can you give me some suggestion? Any word would be helpful!
Best wishes!
File "/home/galen/deepLearningCode/PoseEstimation/G2L_Net/utils/utils_funs.py", line 239, in showpoints
dll = np.ctypeslib.load_library('../utils/render_balls_so', '.')
It seems that this file is not provided.
Can you upload this?
Thanks!
Thank you for your excellent work and sharing! Could you share the code of drawing 3D frame for the target object. Thank you.
can not find it.
您好,感谢您优秀的工作。
请问您是如何插入论文中的图片的呢?我尝试了一些方法但都没有您的论文中那么清晰。
The structure of datasets folder should look like ?
作者您好,我们对您的工作非常感兴趣,但是我们在2080 上大概要train30来天(估计)。请问是我们有哪里做错了吗,或者有什么加快训练的方法吗?期待收到您的回复!
Your efforts are great!!
An error occurred when I prepared the execution environment and executed the sample script as shown below. I followed your tutorial and referred to https://github.com/ultralytics/yolov3 but did not find the file "best_1.pt". Would you please tell me if you make it yourself or have a procurement method from somewhere?
$ cd demo
$ python3 test_linemod.py
Traceback (most recent call last):
File "test_linemod.py", line 249, in <module>
model, classifier, classifier_ce, classifier_box, classifier_box_gan, classifier_box_vec, pc, opt, OR,temp = load_models_yolo(obj)
File "test_linemod.py", line 100, in load_models_yolo
model.load_state_dict(torch.load(weights)['model'])
File "/home/b920405/.local/lib/python3.6/site-packages/torch/serialization.py", line 419, in load
f = open(f, 'rb')
FileNotFoundError: [Errno 2] No such file or directory: '../models/1/best_1.pt'
The following Google Drive had a "404 error" and could not be referenced.
# yolov3 pytorch weights
https://drive.google.com/drive/folders/1uxgUBemJVw9wZsdpboYbzUN4bcRhsuAI
Thank you.
In train_G2L.py, there are not codes about train_yolov3 to recognize "cat". Can you provide that code?
Can you provide the code?
In the path/G2L_Net/train/train_G2L.py,we can see a function import form utils.utils_funs ,which named "get_RT_bat".But there aren`t a function named it in untils_funs.
Hello, Thank you for your work!
I'm a little confused about section3.2 Translation localization. You train a teo PointNets to perform 3D segmentation and output the residual distance ||T- T^-||. What' s the mean val T^- of the segmented points means? Is it to calculate the mean value of these points in the camera coordinate system? And how to understand object translation T here?
Thank you for your reply!
If I want to estimate the object position in real time during the practical process, what depth camera is recommended, I use realsense D435 to obtain the depth map and point cloud quality is very bad, thank you for your reply!
Hi, @DC1991 ! Thanks for your excellent work and sharing. Now I'm confused about the training process, and I'm looking forward to your code or descriptions about the training process.
From my observation, the training contains Translation localization, Rotation localization and others. Are these components trained jointly with the main architecture? Or trained separately?
Any description benefits me !
Once again, thanks a lot!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.