Switch branches/tags
Find file History
tomepetty and nealvis fix prototxt filename. added merge_bn to generate non-batchnorm versi…
…on of caffe weight and prototxt (#195)

Signed-off-by: tomepetty <toemayvang86@gmail.com>
Latest commit 6976b89 Oct 22, 2018

README.md

Caffe Networks for the NCSDK

This directory contains multiple subdirectories. Each subdirectory contains software, data, and instructions that pertain to using a specific Caffe neural network with a Neural Compute device such as the Intel Neural Compute Stick. Typically examples are provided that show how the NCSDK can be used compile the network to a graph file and also how to create a program that uses that graph file for inferencing. The sections below are categorized by network type and include a brief explaination of each network.

Caffe Image Classification Networks for NCSDK

Image Classification Network Description
AgeNet Network that classifies a face image into age ranges.
AlexNet Network that classifies images based on the 1000 categories described in Large Scale Visual Recognition Challenge 2012 (ILSVRC2012).
GenderNet Network that classifies a face image as male or female.
GoogLeNet BAIR/BLVC GoogleNet is a network based on GoogleNet, the winner of ILSVRC 2014, that classifies images based on the 1000 categories described in Large Scale Visual Recognition Challenge 2012 (ILSVRC2012).
ResNet-18 Deep Residual network that classifies images based on the 1000 categories described in Large Scale Visual Recognition Challenge 2012 (ILSVRC2012).
SqueezeNet Accuracy similar to AlexNet with many fewer parameters and small model size as described int the SqueezeNet paper. Network that classifies images based on the 1000 categories described in Large Scale Visual Recognition Challenge 2012 (ILSVRC2012).

Caffe Object Detection Networks for NCSDK

Object Detection Network Description
SSD_MobileNet MobileNet Single Shot Detector takes an image, detects the 20 PASCAL object classes as specified in the (Visual Object Classes Challenges), their bounding boxes, and classifications.
TinyYolo This Tiny You Only Look Once model is based on tiny-yolo DarkNet model . Given an image, detects the 20 PASCAL object classes as specified in the (Visual Object Classes Challenges), their bounding boxes, and classifications. Requires some post processing of results to narrow down relevant boxes.