search

vacancy / PreciseRoIPooling

Precise RoI Pooling with coordinate gradient support, proposed in the paper "Acquisition of Localization Confidence for Accurate Object Detection" (https://arxiv.org/abs/1807.11590).
MIT License
770 stars 152 forks source link

RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED #40

Closed FightingZhen closed 4 years ago

FightingZhen commented 4 years ago

Hello! I have used your PrPool model into my own framework. I used a pre-trained resnet-50 to extract features from one image, follwed by feeding into PrPool model to generate a vector with size (1, 1, 1024, 1). Another image is also fed into the pre-trained network to extract features, followed by a GAP layer to generate another vector with the same size. The two vectors are then calculated with MSE loss, which is utilized to back-propogate for the gradients of ROI index. However, when I run the program, an error: 'RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED' occured, can you help me to fix this problem? thank you ~

jwma0725 commented 4 years ago

@FightingZhen ， Hi, I also meet the problem, have you solved it？