smallcorgi / Faster-RCNN_TF

Faster-RCNN in Tensorflow
MIT License
2.34k stars 1.12k forks source link

ValueError #317

Open 245003478 opened 6 years ago

245003478 commented 6 years ago

/home/mengfei/Desktop/faster_rcnn/lib/roi_pooling_layer/roi_pooling.so 2018-09-27 09:28:03.472026: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA Tensor("Placeholder:0", shape=(1, 224, 224, 3), dtype=float32) Tensor("conv5_3/conv5_3:0", shape=(1, 14, 14, 512), dtype=float32) Tensor("rpn_conv/3x3/rpn_conv/3x3:0", shape=(1, 14, 14, 512), dtype=float32) Tensor("rpn_cls_score/rpn_cls_score:0", shape=(1, 14, 14, 18), dtype=float32) Tensor("rpn_cls_prob:0", shape=(?, ?, ?, ?), dtype=float32) Tensor("rpn_cls_prob_reshape:0", shape=(?, ?, ?, 18), dtype=float32) Tensor("rpn_bbox_pred/rpn_bbox_pred:0", shape=(1, 14, 14, 36), dtype=float32) Tensor("Placeholder_1:0", shape=(?, 3), dtype=float32) Tensor("conv5_3/conv5_3:0", shape=(1, 14, 14, 512), dtype=float32) Tensor("rois:0", shape=(?, 5), dtype=float32) Tensor("pool_5:0", dtype=float32) ValueError: Shape of a new variable (fc6/weights) must be fully defined, but instead was (?, 4096).