If I understood it right all the algorithms are mask based right except for Faster-Rcnn model, right? Can I use the cascade-mask-rcnn model for just object detection, i.e., to train the model without segmentation annotations?

[CharlesNJ]

If I understood it right all the algorithms are mask based right except for Faster-Rcnn model, right? Can I use the cascade-mask-rcnn model for just object detection, i.e., to train the model without segmentation annotations?

[CharlesNJ]

I think the answer is no after some research. Mask models need masks to train and if you'd like to use just the detection, just use cascade-rcnn.