Suitable FPGA board for implementing neural network (face recognition)

[saidwivedi]

I want to implement face recognition system on FPGA board. The pre-processing is done in Matlab. So the input will be feature matrix extracted from image. Topology details are as follows :-

No of Input Neuron - 250
No of Hidden Neuron - 25
No of Output Neurons - 40

Activation function - tan-sigmoid
Learning Algorithm - Scaled Conjugate Gradient Algorithm

 

[mrflibble]

Which fpga to pick will largely depend on if it will fit or not. And that will depend roundabout 100% on two things:
1 - your base model for neuron. Yes I read the above. Bonus points for posting rough specs at all BTW. The implementation of neuron will largely dictate resource usage.
2 - your interconnect model.

And you are of course using fixed point for the sigmoid. Of course you are, because you are not a silly person like some people that use floating point because a random paper said so. ;-)