search

you359 / Keras-FasterRCNN

keras implementation of Faster R-CNN
MIT License
334 stars 215 forks source link

Exception: 'a' cannot be empty unless no samples are taken #79

Open brunojerkovic opened 3 years ago

brunojerkovic commented 3 years ago

I get an exception on this line in "train_frcnn.py": selected_neg_samples = np.random.choice(neg_samples, C.num_rois - len(selected_pos_samples), replace=True).tolist() But the network continues to train, even though it prints this exception a lot of times. Is this exception ok, and should I leave it to train until the end or? (I don't know if it matters, but I am using images in .tiff format)