06-命令

车道线模型训练命令：

cd /code/Ultra-Fast-Lane-Detect&&rm -rf dataset&&mkdir dataset&&cd scripts&&python jsontodataset.py --json_file /dataset/road/labels --out /code/Ultra-Fast-Lane-Detect/dataset/labels_output&&python generate_to_train.py --base_path /code/Ultra-Fast-Lane-Detect/dataset&&cd ..&&python train.py --data_root /code/Ultra-Fast-Lane-Detect/dataset --epoch 30 --batch_size 8&&python export_model.py --resume /code/Ultra-Fast-Lane-Detect/output/exp/best.pth&&torch-model-archiver --model-name road --version 1.0 --handler handler.py --export-path model-store --serialized-file /code/Ultra-Fast-Lane-Detect/output/exp/road.onnx --extra-file handler_predictor.py --force&&mv /code/Ultra-Fast-Lane-Detect/output/exp/ /model/&&cp /code/Ultra-Fast-Lane-Detect/model-store/road.mar /model/&&cp /code/Ultra-Fast-Lane-Detect/data.yaml /mode

车道线编译命令：

cd /detvm&&rm -rf images&&mkdir images&&find /dataset/ -type f \( -name "*.jpg" -o -name "*.png" -o -name "*.jpeg" \) | head -n 100 | xargs -I {} cp {} /detvm/images/&&python3 example.py --data_path=/detvm/images --onnx_model=/model/exp/road.onnx --model_input1_shape3=288 --model_input1_shape4=800  --model_input1_name=images  --quantize_mean1=123.675 --quantize_mean2=116.28 --quantize_mean3=103.53 --quantize_scale1=58.395 --quantize_scale2=57.12 --quantize_scale3=57.375 --quantize_prof_img_num=0  --netbin_folder_path=/code/tpu&&cp /model/data.yaml /tpumodel&&cp /code/tpu/onnx_squeezenet_v1.1/model.bin /tpumodel&&cp /code/tpu/onnx_squeezenet_v1.1/net.bin /tpumodel

交通标志模型训练命令：

cd /code/yolov5&&python train.py --data /dataset/data.yaml --weights  yolov5s.pt  --epochs 30 --batch-size 8 --lr 0.01 --imgsz 640&&python export.py --imgsz 640 --opset 9 --simplify --weights runs/train/exp/weights/best.pt --include onnx --data /dataset/data.yaml --export_kneron&&mv runs/train/exp/weights/best.onnx runs/train/exp/weights/best_kneron.onnx&&python export.py --imgsz 640 --opset 9 --weights runs/train/exp/weights/best.pt --include onnx --data /dataset/data.yaml --export_tpu&&python export.py --imgsz 640 --weights runs/train/exp/weights/best.pt --include torchscript --export_torchserve&&cd ..&&mv yolov5/runs/train/exp/weights/best.torchscript yolov5/runs/train/exp/weights/best.torchscript.pt&&torch-model-archiver --model-name best --version 1.0 --serialized-file yolov5/runs/train/exp/weights/best.torchscript.pt --handler yolov5_model_handler.py --extra-files /dataset/data.yaml,yolov5/runs/train/exp/weights/best.onnx&&cp /code/yolov5/runs/train/exp/weights/* /model/&&cp /code/dataset/data.yaml /model/&&cp /code/best.mar /model/

交通标志编译命令：

cp /model/data.yaml /tpumodel/&&cd /detvm&&rm -rf images&&mkdir images&&find /dataset/ -type f \( -name "*.jpg" -o -name "*.png" -o -name "*.jpeg" \) | head -n 100 | xargs -I {} cp {} /detvm/images/&&python3 example.py --data_path=/detvm/images --onnx_model=/model/best.onnx --model_input1_shape3=640 --model_input1_shape4=640  --model_input1_name=images --quantize_mean1=0.0 --quantize_mean2=0.0 --quantize_mean3=0.0 --quantize_scale1=255.0 --quantize_scale2=255.0 --quantize_scale3=255.0 --quantize_prof_img_num=0 --netbin_folder_path=/tpumodel