I am using darknet_ros to process a webcam image and detect car/bus/trucks for Emergency Vehicle algorithm detection. I see the bounding box sometimes does not cover the entire object and wondering which improvements can be made.
I am providing a 1920x1080 image to darknet_ros module, is this OK or should I provide a lower resolution one?
Other thing I detected is related to the lights on top of a vehicle being detected by yolo as traffic lights, is there any way to filter/disable traffic light classification?
Hi
I am using darknet_ros to process a webcam image and detect car/bus/trucks for Emergency Vehicle algorithm detection. I see the bounding box sometimes does not cover the entire object and wondering which improvements can be made.
Here is an example: https://vimeo.com/616958745/f5eb15cdbe
I am providing a 1920x1080 image to darknet_ros module, is this OK or should I provide a lower resolution one?
Other thing I detected is related to the lights on top of a vehicle being detected by yolo as traffic lights, is there any way to filter/disable traffic light classification?