search

JialeCao001 / SipMask

SipMask: Spatial Information Preservation for Fast Image and Video Instance Segmentation (ECCV2020)
https://arxiv.org/pdf/2007.14772.pdf
MIT License
334 stars 54 forks source link

Batched inference support #33

Open SamSamhuns opened 3 years ago

SamSamhuns commented 3 years ago

Does the model support batched inference support?

I removed the 1 image per GPU constraint and sent a tensor of shape [32, 3, 704, 1344] into the model (SipMask-mmdetection/mmdet/configs/sipmask/sipmask_r101_caffe_fpn_gn_ms_4x.py), but the output from the model is the same for shapes [32, 3, 704, 1344] and [1, 3, 704, 1344]

JialeCao001 commented 3 years ago

@SamSamhuns I am not sure about batched inference.