CV Tasks 包含Classification、Object Detection、Semantic Segmentation、Instance Segmentation,以下為一些應用範例。
此範例說明如何使用Pre-trained的model來做Face Verification和Face Recognition應用。[1]
此範例說明如何用Pre-trained的Mask RCNN來做Object Detection和Instance Segmentation。[2]
此範例說明如何用TensorFlow內建的Pre-trained model來做Classification的Transfer Learning應用。[3]
此範例說明如何用TensorFlow API訓練自己的Object Detection應用。[4]
此範例說明如何使用Pre-trained的model來訓練自己的Semantic Segmentation應用。[5]
此範例說明如何使用Detectron2(Pre-trained model)來做Object Detection、Instance Segmentation、Keypoint Detection、Panoptic segmentation。
Keypoint Detection為體態辨識。[6]
Panoptic segmentation為instance segmentation和semantic segmentation的結合。在圖中可以被數數量出來的物件 (例:bicycle, dog, car, person)被稱為‘things’,難以被計數的區域(例:pavement, ground, dirt, wall)稱為‘stuff’。[7]
此範例說明如何使用U-Net來訓練自己的Semantic Segmentation應用。[8]
此範例介紹圖片Style Transfer應用,輸入原圖和風格圖即可將原圖改變風格。[9]
此範例說明如何使用Poly-YOLO來訓練自己的Object Detection應用。[10]
此範例說明如何使用YOLOv5來訓練自己的Object Detection應用。[11]
此範例說明如何使用StyleGAN來產生超擬真的人臉,並在fidelity和diversity之間做trade-off。[12][13][14]
此範例說明如何使用BigGAN來產生上百種的物品圖。[15]
[1] https://github.com/nyoki-mtl/keras-facenet
[2] https://github.com/matterport/Mask_RCNN
[3] https://www.tensorflow.org/tutorials/images/transfer_learning
[4] https://github.com/EdjeElectronics/TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10
[5] https://github.com/GeorgeSeif/Semantic-Segmentation-Suite.git
[6] https://github.com/facebookresearch/detectron2
[7] https://medium.com/@danielmechea/what-is-panoptic-segmentation-and-why-you-should-care-7f6c953d2a6a
[8] https://github.com/qubvel/segmentation_models
[9] https://github.com/GlebBrykin/SANET
[10] https://gitlab.com/irafm-ai/poly-yolo
[11] https://github.com/ultralytics/yolov5
[12] https://github.com/NVlabs/stylegan
[13] https://jonathan-hui.medium.com/gan-stylegan-stylegan2-479bdf256299
[14] https://arxiv.org/pdf/1812.04948.pdf
[15] https://tfhub.dev/deepmind/biggan-deep-256/1