LayoutParser used Faster-RCNN and MaskRCNN to do DLA, but the model size are over 300M, and hard to do inference in android. Maybe DLA is similar with object-detection, so can we use yolov5/tiny-yolo to replace Faster-RCNN/MaskRCNN?
54 this is an even better version of yolo, though it is based on paddledetection - You might need to install extra dependencies to have it running. And I'll have it updated in the next few PRs.
LayoutParser used Faster-RCNN and MaskRCNN to do DLA, but the model size are over 300M, and hard to do inference in android. Maybe DLA is similar with object-detection, so can we use yolov5/tiny-yolo to replace Faster-RCNN/MaskRCNN?