Mask RCNN with pytorch and pascal2012 dataset
*you need to train longer to get better result
ubuntu18.04
CUDA Version: 10.2
Driver Version: 440.100
GeForce RTX 2060
pytorch 1.5
This repository supports VOC datasets.
PASCAL VOC 2012 (download): http://host.robots.ox.ac.uk/pascal/VOC/voc2012/
The code will check the dataset first before start, filtering samples without annotations.
sh train.sh
sh inference.sh
Note: This is a simple model and only supports batch_size = 1.