Inference module for custom trained model

[JingjunYi]

Hello, thanks for your great work. After i trained superpoint on my dataset, I can't find some way to extract feature points in my own database. I try to simply load image by using Opencv and calculate pred with those images but get some error'Expected 4-dimensional input for 4-dimensional weight [64, 1, 3, 3], but got 3-dimensional input of size [992, 744, 3] instead'. So i use your dataloader and modify it to Infer.py. And for configuring the path of dataset I change 'path' to a option in yaml. Thanks again for this code, but i still have many problems with your dataloader, it seems that it's not simply loading images. I'll appreciate it if you can tell more details about it.

[JingjunYi]

I fork it at here https://github.com/Blackpinkup/pytorch-superpoint

[Chang-Paris]

Hi @Blackpinkup I also had the same problem, I just wanted to run prediction on a simple image to get key points and descriptors. Have you succeeded digging out the reason of wrong input dimension ? Thank you in advance

[JingjunYi]

I think you can solve your problem by browsing my project https://github.com/Blackpinkup/pytorch-superpoint. @Chang-Paris

[weiningwei]

I think you can solve your problem by browsing my project https://github.com/Blackpinkup/pytorch-superpoint. @Chang-Paris

@Blackpinkup , Can you tell more details about how to predict on a simple image ? In the https://github.com/Blackpinkup/pytorch-superpoint, which file can do this?

[JingjunYi]

I have add a guideline in my own project(modification). Please use export.py(based on Infer.py) to export your final result. @Chang-Paris @weiningwei

[Chang-Paris]

Thank you @Blackpinkup for your scripts, it works for me. @weiningwei , pay attention to your config file and dataset, you may also have to bring some modification when you load parameters. The command line would be like python export.py inference your_config.yaml output_name

[fenglezhou]

@JingjunYi Hello, can you please ask in detail about how to use your own data set for evaluation? If it is convenient, can you contact me by email?

[li575867]

Thank you @Blackpinkup for your scripts, it works for me. @weiningwei , pay attention to your config file and dataset, you may also have to bring some modification when you load parameters. The command line would be like python export.py inference your_config.yaml output_name

两位好，方便提供一下能够直接运行的config吗？

[li575867]

@JingjunYi Hello, can you please ask in detail about how to use your own data set for evaluation? If it is convenient, can you contact me by email?

你好，后面有没有运行出来？方便提供一下能够直接运行起来的配置吗？