周易AIPU 仿真R329开发板申请——模型仿真测试

Posted 2021-09-06 极术社区

首发极术社区
如对Arm相关技术感兴趣，欢迎私信aijishu20加入技术微信群。

1. 原始模型文件
   ResNet V1 50
   http://download.tensorflow.org/models/resnet_v1_50_2016_08_28.tar.gz
2. 矫正集
   dataset.npy和label.npy矫正集文件
   链接: https://pan.baidu.com/s/1n-svVZ45FEXzojZ_a5dz0w 提取码: vadh
3. cfg 文件
   resnet_50_run.cfg
   链接: https://pan.baidu.com/s/1kNZ5Wt0FpGFYXYVSiwIidQ 提取码: 2umb

[Common]
mode=run

[Parser]
model_name = resnet_50
detection_postprocess = 
model_domain = image_classification
output = resnet_v1_50/predictions/Reshape_1
input_model = ./resnet_v1_50_frozen.pb
input = input
input_shape = [1,224,224,3]
output_dir = ./

[AutoQuantizationTool]
model_name = resnet_50
quantize_method = SYMMETRIC
ops_per_channel = DepthwiseConv
calibration_data = ./dataset.npy
calibration_label = ./label.npy
preprocess_mode = normalize
quant_precision=int8
reverse_rgb = False
label_id_offset = 0

[GBuilder]
inputs=./input.bin
simulator=aipu_simulator_z1
outputs=output_resnet_50.bin
profile= True
target=Z1_0701

4. simulator 执行结果
   aipubuild resnet_50_run​.cfg
   运行之后得到 output_resnet_50.bin，使用 quant_predict.py 解析后：

predict first 5 label:
    index  230, prob  11, name: Old English sheepdog, bobtail
    index  232, prob   2, name: collie
    index  224, prob   1, name: schipperke
    index  999, prob   0, name: ear, spike, capitulum
    index  329, prob   0, name: sea urchin
true first 5 label:
    index  230, prob 109, name: Old English sheepdog, bobtail
    index  231, prob  96, name: Shetland sheepdog, Shetland sheep dog, Shetland
    index  232, prob  57, name: collie
    index  226, prob  54, name: malinois
    index  263, prob  53, name: Brabancon griffon
Detect picture save to result.jpeg

测试图片：

5.模型实机测试
待审核通过收到实机后测试

相关阅读：

• 【周易AIPU 仿真】R329开发板系列一：环境的搭建和PC端仿真
• 【周易AIPU 仿真】R329 + ubuntu20.04+ docker+ tf_vgg_16