Dear Nvidia TA:
I try to make customer 300*200 pixels images, those must be resized in order to satisfy the detection model’s input dimension. I was used Nvidia’s DetectNet as our main object detection model in DIGITS v6. But the DetectNet default it’s set to 1392 x 512. I can not find a resolution and i want to train my self’s model based on the KITTI dataset (pig…). How to customize the detection model will be explained? How to automate resizing of hundreds of pictures anyone wrote a simple script use Python code ?
Create train_db DB
ERROR: ValueError: Invalid label format in “/home/my/Downloads/labellmg/Kitty/train/labels/000008.txt”