ziyuguo99 / point-bind_point-llm Goto Github PK
View Code? Open in Web Editor NEWAlign 3D Point Cloud with Multi-modalities for Large Language Models
License: MIT License
Align 3D Point Cloud with Multi-modalities for Large Language Models
License: MIT License
Hi, thanks for sharing your amazing work. After going through your paper and some related work, I have some questions that I hope you could shed some light on. They are mainly about the downstream utilization of Point-Bind.
Question 1: This question is about the Any-to-3D Generation part. In my understanding, Point-Bind does not train the Image-Bind encoder, so does this mean that for text/image/audio-to-3D tasks, your approach is no different from directly applying Image-Bind features to CLIP-Forge's decoder?
Question 2: This question is about the Point-LLM part. It seems to me that during the training procedure, you finetuned LLaMA using the same strategy with ImageBind-LLM. Then in the inference part, you add in features extracted from the Point-Bind encoder. I am a little confused with the difference between Point-LLM and ImageBind-LLM. Also, it seems that in the ImageBind-LLM paper, they mention that for 3D domain instructions, they utilize Point-Bind to encode inputs.
Question 3: This question is also about the Point-LLM part. During inference, you feed the Point-Bind extracted features to the visual cache model to retrieve top-k similar ImageBind-encoded features. While this does address the semantic gap between 2D-3D encoders, doesn't this reduce the task of 3D question answering back to some sort of 2D scene question answering? Just like the example given in your paper, when the Point-LLM is given a point cloud of a plane and asked to describe the details of this object, it provides details about the color which seems unlikely to be learned from the point cloud itself (in my understanding, this might be because the top-k images features has encoded color information).
Thanks in advance and again for sharing your work.
Hi, thanks for sharing your amazing work. I have two questions about the demo you provided.
(1) I noticed that you posted "Try our 💥 [Online Demo] here, which is integrated into [ImageBind-LLM]" and I entered the web link. But it seems the SPHINX-MLLM Demo, and I don’t see the entrance to submit point cloud data. So where can I submit point cloud data and generate corresponding descriptions like in the paper?
(2) Based on the first questions, in addition to the online demo, can you provide the inference code? For example, something like SPHINX(https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX)?
Thanks.
Thanks for the excellent work!
Thanks for sharing the great project! This project associates 3d with more modalities than before thus gains more benefits.
I wonder could you also share the pretrained weights of point-bind? E.g., i2p-mae.pt and pointbert.pt
Thanks.
Thank you for your wonderful work and I would like to ask what kind of GPU you used to train the model and the time spent on training.
I really appreciate the fantastic job done the authors in the project.
When I tried to interact with the LLaMA, I got the following errors:
Traceback (most recent call last):
File "exp.py", line 7, in <module>
model = llama.load("/home/qw/proj/ckpt/llama2/llama-2-7b-chat", adapter_ckpt='/home/qw/proj/Point-Bind_Point-LLM/ckpt/7B.pth', knn=True)
File "/home/qw/proj/Point-Bind_Point-LLM/Point-LLM/llama/llama_adapter.py", line 339, in load
model = LLaMA_adapter(
File "/home/qw/proj/Point-Bind_Point-LLM/Point-LLM/llama/llama_adapter.py", line 53, in __init__
self.tokenizer = Tokenizer(model_path=llama_tokenizer)
File "/home/qw/proj/Point-Bind_Point-LLM/Point-LLM/llama/tokenizer.py", line 17, in __init__
self.sp_model = SentencePieceProcessor(model_file=model_path)
File "/home/qw/.conda/envs/pb/lib/python3.8/site-packages/sentencepiece/__init__.py", line 447, in Init
self.Load(model_file=model_file, model_proto=model_proto)
File "/home/qw/.conda/envs/pb/lib/python3.8/site-packages/sentencepiece/__init__.py", line 905, in Load
return self.LoadFromFile(model_file)
File "/home/qw/.conda/envs/pb/lib/python3.8/site-packages/sentencepiece/__init__.py", line 310, in LoadFromFile
return _sentencepiece.SentencePieceProcessor_LoadFromFile(self, arg)
RuntimeError: Internal: src/sentencepiece_processor.cc(1101) [model_proto->ParseFromArray(serialized.data(), serialized.size())]
I would be grateful if someone could help me with this problem.
In addition, I think it would be great if authors could also specify the desired version of required packages (e.g., pytorch, torchaudio, torchvision and cudatoolkit, etc).
您好:
感谢您开源了代码。我想请教一下,您在example文件夹下提供的.pt文件是已经采样了8192个点,是怎么从原始的点云中提取的呢?谢谢
Hello, first of all thank you for sharing your amazing work
I'm trying to upload a pointcloud to the demo web, but it gets stuck uploading and never finishes. I tried with multiple pointcloud, .ply format.
Could you provide a script or instructions about how to perform 3D Q&A given point cloud?
Thanks
There are airplane.pt, examples/car.pt, and examples/toilet.pt in the example.
But I have never seen this format of point cloud.
How can I test my own point clouds?
There is no requirements.txt in the repo.
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.