Kept following this guide to detect license plates. Since I had the labels I moved on to Google Colab. Don’t forget to set the runtime to a GPU.

I tried to mount Google Drive but got this issue. Apparently I was using “Blue’s Solution”. This is what worked instead, adding a Files tab after I agreed to allow the Colab to connect to my Google Drive.

from google.colab import drive 
drive.mount('/content/gdrive', force_remount=True)

Then just copy the folder path. Note that the Ctrl+c and Ctrl+v shortcuts work despite there being no Paste option (It’s under Edit) in the context menu (what shows up when you right-click).

root_dir = "/content/gdrive/MyDrive/Colab Notebooks/"
base_dir = root_dir + 'Cars/'

It then mentions Darknet, an open source neural network framework which has several reources (1, 2, 3), I assume it’s the open-source version of the Ultralytics YOLO. which I used earlier to detect cars.

!git clone https://github.com/AlexeyAB/darknet # Makes a darknet folder

Now set GPU, CUDNN and OPENCV to 1, this is easy to do with Python.

# https://stackoverflow.com/questions/39086/search-and-replace-a-line-in-a-file-in-python

f = open('/content/darknet/Makefile', 'r')
lines = ''
for line in f:
    if line.find('GPU=0') != -1:
        lines += line.replace('GPU=0', 'GPU=1')
        continue
    if line.find('CUDNN=0') != -1:
        lines += line.replace('CUDNN=0', 'CUDNN=1')
        continue
    if line.find('OPENCV=0') != -1:
        lines += line.replace('OPENCV=0', 'OPENCV=1')
        continue
    lines += line
f.close()

g = open('/content/darknet/Makefile', 'w')
g.write(lines)
g.close()

Then we compile Darknet.

!cd "/content/darknet";make --silent;clear;echo "Darknet Compiled!"

Note that anything starting with a bang (!) is a shell command and not Python so a print is equal to an !echo "Darknet Compiled.", very smooth Google Jupyter. Now let’s configure YOLO itself.

!cp /content/darknet/cfg/yolov3.cfg /content/darknet/cfg/yolov3-train.cfg

f = open('/content/darknet/cfg/yolov3-train.cfg', 'r')
lines = ''
for line in f:
    if line.find('batch=1') != -1:
        lines += line.replace('batch=1', 'batch=64')
        continue
    if line.find('subdivisions=1') != -1:
        lines += line.replace('subdivisions=1', 'subdivisions=16')
        continue
    if line.find('max_batches=500200') != -1:
        lines += line.replace('max_batches=500200', 'max_batches=2000')
        continue
    if line.find('filters=255') != -1:
        lines += line.replace('filters=255', 'filters=18')
        continue
    if line.find('classes=80') != -1:
        lines += line.replace('classes=80', 'classes=1')
        continue
    lines += line
f.close()

And make two new files.

!echo -e 'license-plate' > /content/darknet/data/obj.names 
!echo -e 'classes = 1\ntrain = /content/darknet/data/train.txt\nvalid = /content/darknet/data/test.txt\nnames = /content/darknet/data/obj.names\nbackup = /content/yolo-license-plates' > /content/darknet/data/obj.data

Now we unzip the data.

!mkdir data/obj !unzip ../images.zip -d data/obj

Get the YOLO weights.

!cd /content/darknet;wget https://pjreddie.com/media/files/darknet53.conv.74

And use Darknet.

!cd *content/darknet;./darknet detector train data/obj.data cfg/yolov3-train.cfg darknet53.conv.74 -dont_show

To be continued.