road_connectivity
road_connectivity copied to clipboard
anilbatra2185
share .pth.tar file
Thank you very much for your work! I trained the deepglobe dataset with your train_mtl.py , but the prediction is not very good, even worse than using segmentation network like unet or linknet. Could you share your model file? And other question,How the orientation task enhance the final result,add two tasks result?Thanks a lot!
In case you are interested in road detection methods: https://github.com/ntelo007/road_detection_mtl
In case you are interested in road detection methods: https://github.com/ntelo007/road_detection_mtl
Thank you so much for your work! I find it really hard to train the model whose IOU is as high as the paper showed, usually less than 5%. I wonder could you share the model file 'xxx.pth', thanks a lot!
@XinlingQiu I have the same problem with you. The IOU is 5-6% less than that in the paper. Did you solve the question?
Hi @XinlingQiu, @kangkau
Thank you for using the work to explore your research. I assume you are using the similar splits for train/valid on DeepGlobe (available in the repo: link). I have reported the number on the valid set.
Regards Anil
Thank you for your advice! I used the same splits for train/valid on DG, but the result on DG homepage is still less than 0.6. So I wonder if your iou result is derived from your own splited valid set.
Hi @kangkau, The IoU mentioned in the paper is on my splits which are available in the repo.
Thanks Anil
How can I fuse the results of the orientation task and the segmentation task to enhance the final result? Thank you! Looking forward to your reply!
