Image classfication
Easy-to-use/Easy-to-deploy/Easy-to-develop
| *** | example | |
|---|---|---|
| models | (efficientnet/mobilenet/resnest/seresnext等) | 1 |
| metric | (Swish/ArcMarginProduct_subcenter/ArcFaceLossAdaptiveMargin/...) | 2 |
| data aug | (rotate/flip/...、mixup/cutmix) | 3 |
| loss | (ce_loss/ce_smothing_loss/focal_loss/bce_loss/...) | 4 |
| deploy | (flask/grpc/BentoML等) | 5 |
| onnx/trt | () | 6 |
models：
RESNEST_LIST = ['resnest50', 'resnest101', 'resnest200', 'resnest269']
SERESNEXT_LIST = ['seresnext101']
GEFFNET_LIST = ['GenEfficientNet', 'mnasnet_050', 'mnasnet_075', 'mnasnet_100', 'mnasnet_b1', 'mnasnet_140', 'semnasnet_050', 'semnasnet_075', 'semnasnet_100', 'mnasnet_a1', 'semnasnet_140', 'mnasnet_small','mobilenetv2_100', 'mobilenetv2_140', 'mobilenetv2_110d', 'mobilenetv2_120d', 'fbnetc_100', 'spnasnet_100', 'efficientnet_b0', 'efficientnet_b1', 'efficientnet_b2', 'efficientnet_b3', 'efficientnet_b4', 'efficientnet_b5', 'efficientnet_b6', 'efficientnet_b7', 'efficientnet_b8', 'efficientnet_l2', 'efficientnet_es', 'efficientnet_em', 'efficientnet_el', 'efficientnet_cc_b0_4e', 'efficientnet_cc_b0_8e', 'efficientnet_cc_b1_8e', 'efficientnet_lite0', 'efficientnet_lite1', 'efficientnet_lite2', 'efficientnet_lite3', 'efficientnet_lite4', 'tf_efficientnet_b0', 'tf_efficientnet_b1', 'tf_efficientnet_b2', 'tf_efficientnet_b3', 'tf_efficientnet_b4', 'tf_efficientnet_b5', 'tf_efficientnet_b6', 'tf_efficientnet_b7', 'tf_efficientnet_b8', 'tf_efficientnet_b0_ap', 'tf_efficientnet_b1_ap', 'tf_efficientnet_b2_ap', 'tf_efficientnet_b3_ap', 'tf_efficientnet_b4_ap', 'tf_efficientnet_b5_ap', 'tf_efficientnet_b6_ap', 'tf_efficientnet_b7_ap', 'tf_efficientnet_b8_ap', 'tf_efficientnet_b0_ns', 'tf_efficientnet_b1_ns', 'tf_efficientnet_b2_ns', 'tf_efficientnet_b3_ns', 'tf_efficientnet_b4_ns', 'tf_efficientnet_b5_ns', 'tf_efficientnet_b6_ns', 'tf_efficientnet_b7_ns', 'tf_efficientnet_l2_ns', 'tf_efficientnet_l2_ns_475', 'tf_efficientnet_es', 'tf_efficientnet_em', 'tf_efficientnet_el', 'tf_efficientnet_cc_b0_4e', 'tf_efficientnet_cc_b0_8e', 'tf_efficientnet_cc_b1_8e', 'tf_efficientnet_lite0', 'tf_efficientnet_lite1', 'tf_efficientnet_lite2', 'tf_efficientnet_lite3', 'tf_efficientnet_lite4', 'mixnet_s', 'mixnet_m', 'mixnet_l', 'mixnet_xl', 'tf_mixnet_s', 'tf_mixnet_m', 'tf_mixnet_l', 'mobilenetv3_rw', 'mobilenetv3_large_075', 'mobilenetv3_large_100', 'mobilenetv3_large_minimal_100','mobilenetv3_small_075', 'mobilenetv3_small_100', 'mobilenetv3_small_minimal_100','tf_mobilenetv3_large_075', 'tf_mobilenetv3_large_100', 'tf_mobilenetv3_large_minimal_100','tf_mobilenetv3_small_075', 'tf_mobilenetv3_small_100', 'tf_mobilenetv3_small_minimal_100']
train/test/deploy
0、Data format transform
git clone https://github.com/MachineLP/PyTorch_image_classifier
pip install -r requirements.txt
cd PyTorch_image_classifier
python tools/data_preprocess.py --data_dir "./data/data.csv" --n_splits 5 --output_dir "./data/train.csv" --random_state 2020
resnest101
1、Modify configuration file
cp conf/resnest101.yaml conf/resnest101.yaml
vim conf/resnest101.yaml
2、Train:
python train.py --config_path conf/resnest101.yaml
3、Test
python test.py --config_path "conf/resnest101.yaml" --n_splits 5
4、Infer
python infer.py --config_path "conf/resnest101.yaml" --img_path "./data/img/0male/0(2).jpg" --fold "0"
pre>>>>> [1]
python infer.py --config_path "conf/resnest101.yaml" --img_path "./data/img/1female/1(5).jpg" --fold "1"
pre>>>>> [0]
5、Models transform ( https://github.com/NVIDIA-AI-IOT/torch2trt )(Tensorrt installation guide on Ubuntu1804)
onnx：python tools/pytorch_to_onnx.py --config_path "conf/resnest101.yaml" --img_path "./data/img/0male/0(2).jpg" --batch_size 4 --fold 0 --save_path "lp.onnx"
'''
load model ok.....
>>>>> [[-0.15416172 0.36190417]]
cost time: 0.050855159759521484
==> Exporting model to ONNX format at 'lp.onnx'
>>>>> (1, 3, 512, 512)
preds>>>>> [array([[-0.15416166, 0.36190417]], dtype=float32)]
cost time: 3.649467706680298
error_distance: 2.9802322e-08
'''
tensorrt：python tools/onnx_to_tensorrt.py --config_path "conf/resnest101.yaml" --img_path "./data/img/0male/0(2).jpg" --batch_size 4 --fold 0 --save_path "lp_pp.onnx" --trt_save_path "lp.trt"
'''
outputs: tensor([[-0.1543, 0.3619]])
tensor([0.6263]) tensor([1])
'''
6、Deploying models serving
effb3_ns
1、Modify configuration file
cp conf/test.yaml conf/effb3_ns.yaml
vim conf/effb3_ns.yaml
2、Train:
python train.py --config_path "conf/effb3_ns.yaml"
3、Test
python test.py --config_path "conf/effb3_ns.yaml" --n_splits 5
4、Infer
python infer.py --config_path "conf/effb3_ns.yaml" --img_path "./data/img/0male/0(2).jpg" --fold "0"
pre>>>>> [1]
python infer.py --config_path "conf/effb3_ns.yaml" --img_path "./data/img/1female/1(5).jpg" --fold "1"
pre>>>>> [0]
5、Models transform ( https://github.com/NVIDIA-AI-IOT/torch2trt )(Tensorrt installation guide on Ubuntu1804)
onnx：python tools/pytorch_to_onnx.py --config_path "conf/effb3_ns.yaml" --img_path "./data/img/0male/0(2).jpg" --batch_size 4 --fold 0 --save_path "lp.onnx"
tensorrt：python tools/onnx_to_tensorrt.py
6、Deploying models serving
ref
（1）https://github.com/haqishen/SIIM-ISIC-Melanoma-Classification-1st-Place-Solution
（2）https://github.com/BADBADBADBOY/pytorchOCR
（3）https://github.com/MachineLP/QDServing
（4）https://github.com/bentoml/BentoML
（5）mixup-cutmix:https://blog.csdn.net/u014365862/article/details/104216086
（7）focalloss:https://blog.csdn.net/u014365862/article/details/104216192
（8）https://blog.csdn.net/u014365862/article/details/106728375 / https://blog.csdn.net/u014365862/article/details/106728402