MaskRCNN avec tf 2.3
-
conda create -n maskRCNN python=3.8
-
conda install opencv
-
pip install numpy scikit-image tensorflow-gpu keras h5py imgaug
-
conda activate maskRCNN
-
conda install jupyter notebook
-
cd PATH-DE-LA-DEMO
-
git clone https://github.com/pdollar/coco.git
-
cd coco/PythonAPI
-
python setup.py install
-
(si erreur rencontré mettre à jour https://docs.microsoft.com/en-us/answers/questions/136595/error-microsoft-visual-c-140-or-greater-is-require.html)
-
jupyter notebook
Démo dans samples/Mask-RCNN-Demo.ipynb, permet le téléchargement du modèle pré-entrainé coco.
@misc{matterport_maskrcnn_2017, title={Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow}, author={Waleed Abdulla}, year={2017}, publisher={Github}, journal={GitHub repository}, howpublished={\url{https://github.com/matterport/Mask_RCNN}}, }