I am trying to find most efficient and robust Object detector+Segmentation model. I came to know about Mask-rcnn, Yolov5, Yolact, yolov7. As, YOlov7 is new and i read somewhere that yolov7 surpasses all object detectors. Is it true ? If yes kindly elaborate how and let me know your suggestion too that Is there any detector similar to mrcnn or yolov7 available that crosses both ?. I want that model should properly place the mask on object and correctly identify it .
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
As far as I am concerned, Mask R-CNN is the oldest model from listed above.
It comes from R-CNN family, these models are two stage models. Generally speaking, first they make region proposal and then classify them,
Yolo family is younger, models from this family are single stage networks, they spit image into grid and return probabilty of classification.
Besides of that, compatibilty of mask r-cnn with e.g tensorfow 2.0 is problematic. As you mentioned yolo models are faster. Taking into consideration argumemsts above I would choose one from yolo family e.g yolo v7
