search

Yuliang-Liu / Curve-Text-Detector

This repository provides train＆test code, dataset, det.&rec. annotation, evaluation script, annotation tool, and ranking.
639 stars 155 forks source link

train #33

Closed ffyAI closed 5 years ago

ffyAI commented 5 years ago

Hello, I made my own dataset in the same way as the ctw1500 dataset. I encountered the following error during training. How should I train my own dataset? Thank you!

I0506 12:51:41.890240 5382 net.cpp:893] Ignoring source layer prob Traceback (most recent call last): File "./tools/train_net.py", line 118, in max_iters=args.max_iters) File "/home/user/.conda/envs/yffCaffe/Curve-Text-Detector-master/tools/../lib/ pretrained_model=pretrained_model) File "/home/user/.conda/envs/yffCaffe/Curve-Text-Detector-master/tools/../lib/ self.solver.net.layers[0].set_roidb(roidb) File "/home/user/.conda/envs/yffCaffe/Curve-Text-Detector-master/tools/../lib/ self._shuffle_roidb_inds() File "/home/user/.conda/envs/yffCaffe/Curve-Text-Detector-master/tools/../lib/ widths = np.array([r['width'] for r in self._roidb]) KeyError: 'width'

zhangshuaitao commented 5 years ago

@ffyAI, please strictly ensure that your data set is consistent with ctw1500.