MagicSource

@JialeCao001 this speed is much more slower than MaskRCNN when input resolution is considablely large, such as input is 1280 * 800

@JialeCao001 @NaeemKhan333 @cxiang26 His problem is not related with this issue. My problem is the speed shrink down exponentially when input size become larger, and with same large input, MaskRCNN...

@NaeemKhan333 IDK, it's slow when you put the input larger, much more slower than same size input of MaskRCNN

@NaeemKhan333 I using 1280x720, and it's 400ms per image, too slow to say it's as fast as yolact

Configs changed for 2 GPUs: ``` data = dict( samples_per_gpu=2, workers_per_gpu=2, train=dict( type=dataset_type, ann_file=data_root + 'annotations/instances_train2017.json', img_prefix=data_root + 'train2017/', pipeline=train_pipeline), val=dict( type=dataset_type, ann_file=data_root + 'annotations/instances_val2017.json', img_prefix=data_root + 'val2017/', pipeline=test_pipeline), test=dict(...

same here

How to fix psroi cpp build anyway?

选中之后,没有反映,无法输入


@grimoire I will have a test