深度學習在影象領域的應用

隨著深度學習近幾年的火熱發展，在計算機視覺，影象理解方向上，應用越來越廣泛。我們總結了在視覺領域的一些方向上基於深度學習的優秀演算法。包括物體檢測、物體識別、人臉世界、分割、跟蹤、邊緣檢測、影象復原（去雨、去霧）、影象編輯等。

檢測

1. 單一物體檢測

MTCNN: https://github.com/kpzhang93/MTCNN_face_detection_alignment

Cascade-CNN: https://github.com/anson0910/CNN_face_detection

2. 通用物體檢測

Faster-RCNN:

YOLO: https://github.com/pjreddie/darknet

SSD: https://github.com/FreeApe/VGG-or-MobileNet-SSD

RetinaNet: https://github.com/fizyr/keras-retinanet

分類

VGG: https://github.com/ry/tensorflow-vgg16

GoogLenet: https://github.com/n3011/Inception_v3_GoogLeNet

Resnet: https://github.com/ry/tensorflow-resnet

Mobilenet: https://github.com/shicai/MobileNet-Caffe

Shufflenet: https://github.com/anlongstory/ShuffleNet_V2-caffe

MNasNet: https://github.com/zeusees/Mnasnet-Pretrained-Model

識別

1. 人臉識別

Deepface: https://github.com/RiweiChen/DeepFace

Normface: https://github.com/happynear/NormFace

Insightface: https://github.com/deepinsight/insightface

2. 文字識別

DeepOCR: https://github.com/JinpengLI/deep_ocr

跟蹤

1.

2.

分割

Unet: https://github.com/zhixuhao/unet

mask-rcnn: https://github.com/matterport/Mask_RCNN

邊緣檢測

HED: https://github.com/s9xie/hed

RCF: https://github.com/yun-liu/rcf

影象復原

1. 去雨

DDN: https://github.com/XMU-smartdsp/Removing_Rain

CGAN: https://github.com/hezhangsprinter/ID-CGAN

DID-MDN: https://github.com/hezhangsprinter/DID-MDN

DeRaindrop: https://github.com/rui1996/DeRaindrop

2. 去霧

MSCNN: https://github.com/dishank-b/MSCNN-Dehazing-Tensorflow

DehazeNet: https://github.com/caibolun/DehazeNet

3. 超解析度

SRCNN: https://github.com/tegg89/SRCNN-Tensorflow

EDSR: https://github.com/thstkdgus35/EDSR-PyTorch (https://blog.csdn.net/xjp_xujiping/article/details/81986020)

4.影象單反化

DPED: https://github.com/aiff22/DPED