l-bat

import numpy as np
import os
import onnx
import torch
import torch.nn as nn
import torch.nn.functional as F
import cv2

# Class for RPN
class RPN(nn.Module):

l-bat

def xcorr_depthwise(x, kernel):
    """
    Deptwise convolution for input and weights with the same shapes
    Elementwise multiplication -> GlobalAveragePooling -> scalar mul on (kernel_h * kernel_w)
    """
    # batch = kernel.size(0) # in our model batch = 1
    # channel = kernel.size(1)
    # x = x.view(1, batch*channel, x.size(2), x.size(3)) # batch already = 1
    # kernel = kernel.view(batch*channel, 1, kernel.size(2), kernel.size(3))

l-bat

        else if (layer_type == "Gather")
        {
            CV_Assert(node_proto.input_size() == 2);
            Mat indexMat = getBlob(node_proto, constBlobs, 1);
            CV_Assert_N(indexMat.type() == CV_32S, indexMat.total() == 1);
            int index = indexMat.at<int>(0);

            if ((constBlobs.find(node_proto.input(0)) != constBlobs.end()))
            {
                Mat input = getBlob(node_proto, constBlobs, 0);

l-bat

    #ifdef HAVE_DNN_NGRAPH
    virtual Ptr<BackendNode> initNgraph(const std::vector<Ptr<BackendWrapper> > &inputs,
                                        const std::vector<Ptr<BackendNode> >& nodes) CV_OVERRIDE
    {
        auto& input = nodes[0].dynamicCast<InfEngineNgraphNode>()->node;

        auto parent_shape = input->get_shape();

        std::cout << "parent_shape " << parent_shape << std::endl;

l-bat

#include <iostream>
#include <vector>
#include <random>
#include <inference_engine.hpp>
#include <ngraph/ngraph.hpp>
#include <ngraph/opsets/opset.hpp>

using namespace InferenceEngine;

l-bat

        if (inputs.size() == 2)
        {
            int dims = outputs[0].dims;
            int m = inputs[0].size[dims - 2];
            int n = inputs[0].size[dims - 1];
            int k = inputs[1].size[dims - 1];
            int rows = inputs[0].total() / (m * n);

            MatShape sh_A = shape(rows, m * n);
            MatShape sh_B = shape(rows, n * k);

l-bat

import os
import argparse
import caffe
import numpy as np


parser = argparse.ArgumentParser()
parser.add_argument('--proto_dir',  '-i', default='proto', help='Path to directory with prototxt.')
parser.add_argument('--output_dir', '-o', default='output', help='Path to save directory.')
args, _ = parser.parse_known_args()

l-bat

#include <iostream>
#include <vector>
#include <random>
#include <inference_engine.hpp>
#include <ngraph/ngraph.hpp>
#include <ngraph/opsets/opset.hpp>

using namespace InferenceEngine;

l-bat

class FeatureL2Norm(torch.nn.Module):
    def __init__(self):
        super(FeatureL2Norm, self).__init__()

    def forward(self, feature):
        norm = torch.norm(feature, p=2, dim=1, keepdim=True)
        return torch.div(feature, norm)

l-bat

date,price,txCount,txVolume,activeAddresses,symbol,name,open,high,low,close,volume,market
2013-04-28,135.3,41702.0,68798683.4463,117984.0,BTC,Bitcoin,135.3,135.98,132.1,134.21,0.0,1500520000.0
2013-04-28,4.3,9174.0,44319520.0666,17216.0,LTC,Litecoin,4.3,4.4,4.18,4.35,0.0,73773400.0
2013-04-29,134.44,51602.0,113812845.38,86925.0,BTC,Bitcoin,134.44,147.49,134.0,144.54,0.0,1491160000.0
2013-04-29,4.37,9275.0,36478096.6026,18395.0,LTC,Litecoin,4.37,4.57,4.23,4.38,0.0,74952700.0
2013-04-30,144.0,47450.0,84266323.8547,76871.0,BTC,Bitcoin,144.0,146.93,134.05,139.0,0.0,1597780000.0
2013-04-30,4.4,9099.0,40391660.1041,17810.0,LTC,Litecoin,4.4,4.57,4.17,4.3,0.0,75726800.0
2013-05-01,139.0,55176.0,120682532.517,83564.0,BTC,Bitcoin,139.0,139.89,107.72,116.99,0.0,1542820000.0
2013-05-01,4.29,8990.0,76374202.7997,16991.0,LTC,Litecoin,4.29,4.36,3.52,3.8,0.0,73901200.0
2013-05-02,116.38,55295.0,93375328.6292,81920.0,BTC,Bitcoin,116.38,125.6,92.28,105.21,0.0,1292190000.0