chengchunlei
47 lines
1.9 KiB
Markdown
47 lines
1.9 KiB
Markdown
## 1.Install requirements：
|
||
```
|
||
pip3.7.5 install python-opencv
|
||
cd backend_C++/dnmetis_backend
|
||
pip3.7.5 setup.py install
|
||
```
|
||
Details of dnmetis_backend installation can be found in backend_C++/dnmetis_backend/README.md. Notice that, you just need to install requirements once for a brand new Ai1-Inference environment。
|
||
|
||
Here is an example of how to run npu inference of efficientnet-b8:
|
||
## 2.Download model(.om)
|
||
|
||
1.download efficientnet-b8 model(.om): \
|
||
URL：[baidu pan](https://pan.baidu.com/s/1N-kpQoDe3NRxvjFKjAT9AA) \
|
||
Extracted code：tvg0
|
||
|
||
Original tensorflow model of efficientnet-b8(.pb):\
|
||
URL：[baidu pan](https://pan.baidu.com/s/1CajdSlNTh6k35RoyOn-3Ug)\
|
||
Extracted code：slqm \
|
||
|
||
If you want to acknowledge how to generate om from pb，pls download efficientnet-b8.pb and execute ATC cmd：\
|
||
.atc --model=MODELDIR/efficientnet−b8.pb −−framework=3 −−inputshape=′images:1,672,672,3′−−output=MODELDIR/efficientnet-b8 --mode=0 --out_nodes='Softmax:0' --soc_version=Ascend310 --input_fp16_nodes=images --output_type=FP16
|
||
|
||
2.Imagenet-val dataset and labels in val_map.txt:
|
||
|
||
|
||
|
||
|
||
|
||
## 3.Start execute the inference:
|
||
|
||
sh run_efficientnet-b8.sh
|
||
|
||
or
|
||
|
||
python3.7 main.py --model=./model/efficientnet-b8.om --image_size='672,672,3' --inputs='images:0' --outputs='Softmax:0' --precision=fp16
|
||
|
||
|
||
## 4.ATC offline model generate (optional):
|
||
|
||
1.download efficientnet-b8 model(.pb) URL: obs://hwwheel23/efficientnet-b8.pb
|
||
|
||
2.atc --model=$MODEL_DIR/efficientnet-b8.pb --framework=3 --input_shape='images:1,672,672,3' --output=$MODEL_DIR/efficientnet-b8 --mode=0 --out_nodes='Softmax:0' --soc_version=Ascend310 --input_fp16_nodes=images --output_type=FP16
|
||
|
||
## 5.Imagenet2012-val Top1 Accuracy:
|
||
|
||
|