毕设&课程作业_基于Keras预训练模型VGG16等，使用Python的HTTP框架Flask搭建图像识别接口.zip

# Keras-ImageNet 基于Keras预训练模型VGG16、ResNet50、InceptionV3，使用Python的HTTP框架Flask搭建图像识别接口。 > * Download from Baidu Netdisk VGG16、ResNet50、InceptionV3 downloaded: http://pan.baidu.com/s/1geHmOpH details: http://keras-cn.readthedocs.io/en/latest/other/application/ > * Move to ~/.keras/models folder ①imagenet_class_index.json 30+K ②vgg16_weights_tf_dim_ordering_tf_kernels.h5 527+M ③resnet50_weights_tf_dim_ordering_tf_kernels.h5 98+M ④inception_v3_weights_tf_dim_ordering_tf_kernels.h5 91+M > * Run by Flask Http Framework `python run.py` # local test `nohub python run.py &` # remote deploy note : debug must be False if equals True, ValueError: Tensor:(…) is not an element of this graph ``` if __name__ == '__main__': app.run('0.0.0.0', debug=False) # note : must be False ``` > * Request ``` file 图片 type(file) top 标签数量 type(integer) network 网络 select(vgg/res/inc) language 语言 select(cn/en) ``` > * Response ``` { "result": 0, # 0 success 1 error "tags": [ { "tag_confidence": 0.982, "tag_name": "老虎" }, { "tag_confidence": 0.018, "tag_name": "山猫,虎猫" } ] } ``` ``` # For VGG16 curl curl "127.0.0.1:5000/api" -F"top=5" -F"net=vgg" -F"lang=cn" -F"file=@tiger.jpg" # For RestNet50 curl curl "127.0.0.1:5000/api" -F"top=5" -F"net=res" -F"lang=cn" -F"file=@tiger.jpg" # For InceptionV3 curl curl "127.0.0.1:5000/api" -F"top=5" -F"net=inc" -F"lang=cn" -F"file=@tiger.jpg" # For Http Java see ImageNet.java for more details. ```