search

Owen-Liuyuxuan / visualDet3D

Official Repo for Ground-aware Monocular 3D Object Detection for Autonomous Driving / YOLOStereo3D: A Step Back to 2D for Efficient Stereo 3D Detection
https://owen-liuyuxuan.github.io/papers_reading_sharing.github.io/3dDetection/GroundAwareConvultion/
Apache License 2.0
362 stars 77 forks source link

Change the dataset #12

Closed sjg02122 closed 3 years ago

sjg02122 commented 3 years ago

Hi Thanks to your code, I can learn many things. Thank you.

What I'm curious about is changing data. I want to do the training only left image and the test data right image in a monocular 3D

Which part should I modify to execute the code? I'll be waiting for the reply. Thank you.

Owen-Liuyuxuan commented 3 years ago

I am not sure I fully understand your question. If you are only training with the left images, then you can set "cfg.data.use_right_image=False" in the config file .

sjg02122 commented 3 years ago

The data structure I understand is as follows.

Train dataset : train/Image_2 +Image_3 Test dataset : test/Image_2

The way I want it is as follows.

Traing dataset : train/Image_2 Test dataset : test/Image_3

Image_2 : left image Image_3 : right image

Owen-Liuyuxuan commented 3 years ago

It is rare to test with the right images of the testing set, because the observation angle and the 2D bounding boxes will be different when directly run the evaluation function.

But that is doable.

You can modify KittiMonoTestDataset, following here. Notice how to get data from "read_data" function by setting different "output_dict". Or you can write a new dataset class inheriting KittiMonoTestDataset.

sjg02122 commented 3 years ago

Thanks, I will try your command.