arunm8489

def parse_cfg(config_file):
    file = open(config_file,'r')
    file = file.read().split('\n')
    file =  [line for line in file if len(line)>0 and line[0] != '#']
    file = [line.lstrip().rstrip() for line in file]

    final_list = []
    element_dict = {}
    for line in file:

arunm8489

import numpy as np
import pandas as pd 
import torch
import torch.nn as nn
import cv2
import os
import time
from torch.autograd import Variable
import matplotlib.pyplot as plt

arunm8489

config_file = 'yolov3.cfg'
blocks = parse_cfg(config_file)
darknet_details = blocks[0]
channels = 3 
#list of filter numbers in each layer.It is useful while defining number of filters in routing layer
output_filters = []  
modulelist = nn.ModuleList()

arunm8489

   for i,block in enumerate(blocks[1:]):
        seq = nn.Sequential()
        #upcoming code comes under this loop

arunm8489

if (block["type"] == "convolutional"):
         activation = block["activation"]
         filters = int(block["filters"])
         kernel_size = int(block["size"])
         strides = int(block["stride"])
         use_bias= False if ("batch_normalize" in block) else True
         pad = (kernel_size - 1) // 2
         conv = nn.Conv2d(in_channels=channels, out_channels=filters, kernel_size=kernel_size, 
                             stride=strides, padding=pad, bias = use_bias)
            

arunm8489

elif (block["type"] == 'route'):
     # start and end is given in format (eg:-1 36 so we will find layer number from it.
     # we will find layer number in negative format
     # so that we can get the number of filters in that layer
     block['layers'] = block['layers'].split(',')
     start = int(block['layers'][0])
     if len(block['layers']) == 1:               
          #ie if -1 given and present layer is 20 . we have to sum filters in 19th and 20th layer 
          block['layers'][0] = int(i + start)             
          filters = output_filters[block['layers'][0]]  #start layer number

arunm8489

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

arunm8489

elif block["type"] == "yolo":
     mask = block["mask"].split(",")
     mask = [int(m) for m in mask]
     anchors = block["anchors"].split(",")
     anchors = [(int(anchors[i]), int(anchors[i + 1])) for i in range(0, len(anchors), 2)]
     anchors = [anchors[i] for i in mask]
     block["anchors"] = anchors
            
     detectorLayer = DetectionLayer(anchors)
     seq.add_module("Detection_{0}".format(i),detectorLayer)

arunm8489

class DetectionLayer(nn.Module):
    def __init__(self, anchors):
        super(DetectionLayer, self).__init__()
        self.anchors = anchors

arunm8489

modulelist.append(seq)
output_filters.append(filters)     
channels = filters