Faster RCNN model makes only 100 predictions on every image. How to make the model predict dorrectly when the image has more than 100 objects?

1

I use Faster RCNN Predictor

My build model function

1
1718×128 14.8 KB

predictions = model(imgs)

predictions[0][“boxes”].len()
100

2

It seems you might want to change the box_detections_per_img argument in FasterRCNN.