Real-time-Vehicle-and-Pedestrian-Counting-CenterNet
Real-time-Vehicle-and-Pedestrian-Counting-CenterNet copied to clipboard
No custom class detected
Hello Clement,
Thank you for adapting deepsort center net for custom classes. However, for me this failed.
I trained cocodl2x model on coco subset dataset and added 1 custom category (as i have only one class). While testing, the video generates no prediction on this custom category.
Also, when i tried to evaluate over the test records(to see the overall MAP performance), the coco eval script fails for me as the image ids i have are in string format. I made quite a lot of changes in coco eval files but that also failed :(
Any help is appreciated.
Thank you. Best, Anshu