|
问了,没人理,,,我用一张图像量化FPS是6.65,用两百张图像反而成了6.63。用的是python接口,连接1808评价得到的。
- # Config for Model Input PreProcess
- rknn.config(channel_mean_value='127.5 127.5 127.5 127.5', reorder_channel='0 1 2',batch_size = 8)
- #rknn.config(channel_mean_value='0 0 0 1', reorder_channel='0 1 2')
- # Load TensorFlow Model
- print('--> Loading model')
- rknn.load_tensorflow(tf_pb='./model.pb',
- inputs=['lanenet/input_tensor'],
- outputs=['lanenet_model/vgg_frontend/vgg16_decode_module/binary_seg_decode/binary_final_logits/binary_final_logits'],
- input_size_list=[[256, 512, 3]])
- print('done')
- # Build Model
- print('--> Building model')
- rknn.build(do_quantization=True, dataset='./dataset.txt')
- print('done')
- print('--> Begin evaluate model performance')
- perf_results = rknn.eval_perf(inputs=[img]) #评估模型性能。 返回模型在硬件上运行的总时间
- print('done')
|
|