QANet
QANet copied to clipboard
object detection
Hello, does the author do target detection in qanet? Why do we give boxes in target when testing and infence
As far as I know, the code in this repo contains a code for generating the quality score between the network output and target detection. The code doesn`t contain any bounding box detection. You can use an off-the-shelf method such as the openpose to provide the input bounding box (I had tested it and it works fine). Also, you can ignore the quality assessment part if you only want the part segmentation results. I had modified the code and it works fine without the quality part.
As far as I know, the code in this repo contains a code for generating the quality score between the network output and target detection. The code doesn`t contain any bounding box detection. You can use an off-the-shelf method such as the openpose to provide the input bounding box (I had tested it and it works fine). Also, you can ignore the quality assessment part if you only want the part segmentation results. I had modified the code and it works fine without the quality part.
Could you provide a visual code of segmentation results
@zsn-life What do you mean by visual code?