Comparison with other libraries

Video Presentations

• Deep Learning in C++: Comparing tiny-dnn, Caffe, Dynet and other libraries with 1D data

Feature Comparison

	Prerequisites	Modeling	Training	Execution	GPU Support	Installing	Windows Support	Pre-Trained Model
tiny-dnn	Nothing(Optional:TBB,OpenMP)	C++	C++	C++	No	Unnecessary	Yes	Yes(via caffe-converter)
caffe	BLAS,Boost,protobuf,glog,gflags,hdf5,(Optional:CUDA,OpenCV,lmdb,leveldbetc)	Config File	C++, Python	C++, Python	Yes	Necessary	Yes	Yes
Theano	Numpy,Scipy,BLAS,(optional:nose,Sphinx,CUDAetc)	Python Code	Python	Python	Yes	Necessary	Yes	No1
TensorFlow	numpy,six,protobuf,(optional:CUDA,Bazel)	Python Code	Python	Python, C++2	Yes	Necessary	Yes	No1
Mxnet	BLAS(optional:CUDA)	C++, Python, R, Julia ...	C++, Python, R, Julia ...	C++4, Python, R, Julia ...	Yes	Necessary3	Yes	Yes(via caffe-converter)

¹ unofficial version is available
² limited C++ API
³ It is rumoured installation is unnecessary with the Mxnet amalgamation Api
⁴ both full API and limited api via amalgamation