This demo was developed using LeFlow, which is an open source tool available at [ Ссылка ]
Read more about it here: [ Ссылка ]
In this demo, training is performed offline using Tensorflow and the hardware is generated using LeFlow. The network is composed of a 2D convolution with 8 channels, followed by a maxpool, followed by another 2D convolution with 16 channels, another maxpool and a final dense layer.
Ещё видео!