PT TO CVIMODEL

    

Geração do .pt

Entre em

https://colab.research.google.com/github/roboflow-ai/notebooks/blob/main/notebooks/train-yolo11-object-detection-on-custom-dataset.ipynb#scrollTo=tdSMcABDNKW-yolo11

E nele, basicamente deves pegar o ZIP (yolov11) gerado no ROBOFLOW, descompactar, instalar o ULTRALYTICS (yolo)

%pip install "ultralytics<=8.3.40" supervision roboflow

# prevent ultralytics from tracking your activity

!yolo settings sync=False

import ultralytics

ultralytics.checks()

..após isto executar o Script

!yolo task=detect mode=train model=yolo11n.pt data=data.yaml epochs=200 imgsz=640

Salve o .pt gerado

Instale o TPU MLIR e ULTRALYTICS em seu PC

docker run --privileged --name recamera -v /workspace -it sophgo/tpuc_dev:v3.1 on /workspace sudo apt-get update sudo apt-get upgrade pip install tpu_mlir[all]==1.7 git clone https://github.com/sophgo/tpu-mlir.git cd tpu-mlir source ./envsetup.sh ./build.sh mkdir model_yolo11n && cd model_yolo11n cp -rf ${REGRESSION_PATH}/dataset/COCO2017 . cp -rf ${REGRESSION_PATH}/image . mkdir Workspace && cd Workspace pip install ultralytics (crie em /dataset/ o nome de uma pasta com as imagens treinadas) (dentro de /image/copie uma imagem das imagens treinadas) ### git clone https://github.com/Seeed-Studio/sscma-example-sg200x.git ### cd sscma-example-sg200x/scripts ### (copy into this folder the best.onnx) ### python export.py --output_names "/model.23/cv2.0/cv2.0.2/Conv_output_0,/model.23/cv3.0/cv3.0.2/Conv_output_0,/model.23/cv2.1/cv2.1.2/Conv_output_0,/model.23/cv3.1/cv3.1.2/Conv_output_0,/model.23/cv2.2/cv2.2.2/Conv_output_0,/model.23/cv3.2/cv3.2.2/Conv_output_0" --dataset ../../../../tpu-mlir/regression/dataset/BUGGIO --test_input ../../../../tpu-mlir/regression/image/Ades_2-4_jpg.rf.4de8403c125c5d16b435a839a3a93780.jpg best.onnx dentro do /workspace copie o best.pt (gerado no Colab) execute yolo export model=best.pt format=onnx imgsz=640,640 será gerado um best.onnx execute model_transform \ --model_name yolo11n \ --model_def best.onnx \ --input_shapes "[[1,3,640,640]]" \ --mean "0.0,0.0,0.0" \ --scale "0.0039216,0.0039216,0.0039216" \ --keep_aspect_ratio \ --pixel_format rgb \ --output_names "/model.23/cv2.0/cv2.0.2/Conv_output_0,/model.23/cv3.0/cv3.0.2/Conv_output_0,/model.23/cv2.1/cv2.1.2/Conv_output_0,/model.23/cv3.1/cv3.1.2/Conv_output_0,/model.23/cv2.2/cv2.2.2/Conv_output_0,/model.23/cv3.2/cv3.2.2/Conv_output_0" \ --test_input ../../../tpu-mlir/regression/image/Ades_2-4_jpg.rf.4de8403c125c5d16b435a839a3a93780.jpg \ --test_result yolo11n_top_outputs.npz \ --mlir yolo11n.mlir execute run_calibration \ yolo11n.mlir \ --dataset ../BUGGIO \ --input_num 100 \ -o yolo11n_calib_table execute model_deploy \ --mlir yolo11n.mlir \ --quantize INT8 \ --quant_input \ --processor cv181x \ --calibration_table yolo11n_calib_table \ --test_input ../../../tpu-mlir/regression/image/Ades_2-4_jpg.rf.4de8403c125c5d16b435a839a3a93780.jpg \ --test_reference yolo11n_top_outputs.npz \ --customization_format RGB_PACKED \ --fuse_preprocess \ --aligned_input \ --model yolo11n_1684x_int8_sym.cvimodel cvimodel se encontra em /workspace/tpu-mlir/model_yolo11n/Workspace \\wsl.localhost\docker-desktop\mnt\docker-desktop-disk\data\docker\volumes\8bd5aab9644ef6f95e9275c42306a33a7313a58d86c9a1a7ab20e4a24f5649aa\_data\tpu-mlir\model_yolo11n\Workspace ### o script do RECAMERA

REF:

Model Conversion Guide | Seeed Studio Wiki
Roboflow: Computer vision tools for developers and enterprises
sscma-example-sg200x/scripts/export.py at c027ed4ea14564b69530a9958150953182443126 · Seeed-Studio/sscma-example-sg200x
Monkey CVI model · Issue #37 · Seeed-Studio/reCamera-OS

Sobre a SMARTCORE

A SMARTCORE FORNECE CHIPS E MÓDULOS PARA IOT, COMUNICAÇÃO WIRELESS, BIOMETRIA, CONECTIVIDADE, RASTREAMENTO E AUTOMAÇÃO. NOSSO PORTFÓLIO INCLUI MODEM 2G/3G/4G/NB-IOT, SATELITAL, MÓDULOS WIFI, BLUETOOTH, GPS, SIGFOX, LORA, LEITOR DE CARTÃO, LEITOR QR CCODE, MECANISMO DE IMPRESSÃO, MINI-BOARD PC, ANTENA, PIGTAIL, BATERIA, REPETIDOR GPS E SENSORES