I download the kitti images from their website and their sizes are about 1248 376. However, I find the "run_mscnn_detection.m" use default image size as "2560 768". The pretrained model "mscnn-8s-768" also use image height of 768. Why do you need to enlarge the image? What are the meaning of the names "mscnn-xs-xxx" in the example folder?
The input size has an important role in Kitti object detection tasks. Please have a look at our paper for details. "mscnn-xs-xxx" means "mscnn-# of proposal branches-height of input images".
Hi Zhaowei,
I download the kitti images from their website and their sizes are about 1248 376. However, I find the "run_mscnn_detection.m" use default image size as "2560 768". The pretrained model "mscnn-8s-768" also use image height of 768. Why do you need to enlarge the image? What are the meaning of the names "mscnn-xs-xxx" in the example folder?
Thanks