search

Banconxuan / RTM3D

The official PyTorch Implementation of RTM3D and KM3D for Monocular 3D Object Detection
MIT License
453 stars 85 forks source link

Inference on own data (different perspective, camera and resulution) #21

Open walzimmer opened 3 years ago

walzimmer commented 3 years ago

Inference on own data with different perspective, camera and resolution:

rtm3d_on_providentia_first_result_no_calibration_no_crop

When cropping the input image:

rtm3d_on_providentia_first_result_no_calibration_cropped

Changing the calibration parameters may help a bit. Is there anything else I can try?

One solution would be to annotate my own data and then to retrain the network.

s0urcer commented 3 years ago

@walzimmer Have you succeeded in solving this? AFAIU there are no separate cameras intrinsic parameters setup in the architecture so the network is trained to work with specific camera only. Am I missing something?

albert0713 commented 3 years ago

Could you tell me what the command is to inference by my own pics, I just follow the readme, but I cant inference my own pics. Thanks