I would like to demo using a single RGB image as input.

[ttm119]

Hello,

I hope this message finds you well. I am currently engaged in a project that involves processing RGB images and I have come across your repository with great interest. If it's not too much trouble, could you possibly provide a demonstration code that uses a single arbitrary RGB image as an input? Such a code would be immensely helpful for my project, and I would be very grateful for your time and assistance.

[HaolinLiu97]

Hi, I am planning to writing codes on inferencing on arbitary images. I will try to finish it within this week.

Best,

[ttm119]

Thank you so much for your prompt response and for taking the time to work on this code. It's great to hear that you are planning to develop functionality for inferencing on arbitrary images. Your efforts will significantly enhance the versatility of the project and will be extremely beneficial for my work. I am looking forward to the update and appreciate your commitment to completing it within the week.

[mert-o]

Hi @HaolinLiu97,
Thank you for your great work!
I am also working for the rgb demo version, do you plan to release it in a short time?

Best

[HaolinLiu97]

Hi, I should be able to finish within this week. I will try my best to squeeze some time to work on this.

[HaolinLiu97]

Hi guys, the demo on arbitrary images is already released (inside ./real_demo). By the way, the limitation of this work is that it is hard to have good performance on arbitrary images due to the limitation of pix3d dataset and the worse generalizability of using local features. If you are interested, my latest work already solved this problem for instance reconstruction from few views by proposing a new dataset and better algorithm. (ps. codes and data of it will be released within this month)

[ttm119]

Hi Haolin,
Thanks for releasing the demo. I'm trying to run it but I'm having trouble understanding the file structure. Could you please provide a download link for the complete file structure or explain the file structure to me?
I would really appreciate your help.

Thanks,

[HaolinLiu97]

@ttm119 Hi, all the required checkpoints can be downloaded here. Unzip it and put it under the checkpoints folder of this repository. For yolov7, and Im3d, just clone them in somewhere, and install them one by one. Then, in the ./real_demo/run_demo.sh script, modify the yolo_dir and im3d_dir to their location. Also, modify the instpifu_dir to where you clone this repository. And modify the sunrgbd_dir to where you downloa and save the sunrgbd_train_test_data

[csyhping]

@HaolinLiu97 , thanks for your great work. It seems there is a required task_id in real_demo. so, actually, this arbitrary image is still one of the images in the dataset? May I ask how can we use an image out of the dataset? Thanks!