Exception: a must be non-empty

[lengyue1996]

I am training my dataset and I came across this problem when training any one has idea toward this?

[zhimengfan1990]

got the same issue. I think this is due to neg_samples is empty. We can fix like this.

			# we need at least one negative sample for sampling 
			if len(list(neg_samples)) == 0:
				continue 

BTW, can you get high accuracy after training on VOC? I test and find the accuracy is very low. It can only detect persons, but the persons' accuracy is also very low.