After trained model with your tutorial. (Using your dic.txt and fsns in data/train/tfexample.record). Then i use the model-checkpoint to infer 2 image. But the output is not right. Please help me!