markhsia / convnetcs

convolution neural network library in C# with Pretrained models based on convnetjs

ConvNetCS

ConvNetCS Library is a c# .net Convolution neural network library along with expermintal deepQlearning implementation, based on Andrej Karpathy Convnetjs. the network can be built from several layers including :

• Full Connected Layer
• Dropout Layer
• Convolution Layer
• Local Response Normalization
• Max Pool Layer
• Maxout Layer (Activation Layer)
• Relu Layer (Activation Layer)
• Sigmoid Layer (Activation Layer)
• Regression Layer (Loss Layer)
• SVM Layer (Loss Layer)
• Softmax Layer (Loss Layer)

Pretrained Models

• VGG16 for imagenet dataset
• VGG16 for PLaces365 dataset

Load pretrained models from PlayWithVGG project and start visulaize features.

Notes!

• run on release mode without debuging to boost the performance.
• all the pretrained models is converted from the original caffe model and the licence of the original author is applied.

License

MIT