1 00:00:02,000 --> 00:00:03,000 Hello everyone. 2 00:00:03,000 --> 00:00:10,000 In this video tutorial we will see how we can use YOLO V8 for license plate detection and recognition. 3 00:00:10,000 --> 00:00:16,000 So as you can see in this output demo video, we will not only detect the license plate, but we will 4 00:00:16,000 --> 00:00:19,000 also read the text on the license plate as well. 5 00:00:19,000 --> 00:00:24,000 So we will do the detection of the license plate, but also we will read the text of the license plate 6 00:00:24,000 --> 00:00:25,000 as well. 7 00:00:25,000 --> 00:00:32,000 So in this video tutorial we will see how we can detect the license plate using the V8 and also how 8 00:00:32,000 --> 00:00:40,000 we can read the text on the license plate using easy OCR so we can fine tune the YOLO V8 model for license 9 00:00:40,000 --> 00:00:41,000 plate detection. 10 00:00:41,000 --> 00:00:46,000 And then we will use OCR to read the text on the license plate. 11 00:00:46,000 --> 00:00:51,000 So this is a complete end to end tutorial, so don't skip any part of the video and do watch the complete 12 00:00:51,000 --> 00:00:52,000 video. 13 00:00:52,000 --> 00:00:54,000 So let's get started. 14 00:00:56,000 --> 00:01:03,000 So first of all, I will just go to the script file over here and I will explain you the complete code. 15 00:01:03,000 --> 00:01:08,000 So let me move move towards the script file and then we will proceed further and I will explain you 16 00:01:08,000 --> 00:01:10,000 the complete code step by step. 17 00:01:11,000 --> 00:01:11,000 Okay. 18 00:01:11,000 --> 00:01:18,000 So so here in front of you, you can see the CoLab notebook file, Google CoLab notebook file over here. 19 00:01:18,000 --> 00:01:23,000 So before running the script, please make sure that you have selected the runtime as GPU. 20 00:01:24,000 --> 00:01:24,000 Okay. 21 00:01:24,000 --> 00:01:29,000 And click on Save then for this project I will be using this license plate dataset. 22 00:01:29,000 --> 00:01:35,000 The dataset is publicly available in Roboflow, so I will share this dataset link with you as well. 23 00:01:35,000 --> 00:01:40,000 So you just need to create an account on roboflow and it's you can assess this dataset for free. 24 00:01:40,000 --> 00:01:43,000 You just don't need to pay anything to assess this dataset. 25 00:01:43,000 --> 00:01:46,000 So the dataset consist of about 300 images. 26 00:01:46,000 --> 00:01:48,000 We have the train validation and test images. 27 00:01:48,000 --> 00:01:51,000 The author of this dataset has applied augmentation. 28 00:01:51,000 --> 00:01:57,000 So after the data augmentation, the training set is of the images increased to 6 to 18 images. 29 00:01:57,000 --> 00:02:03,000 Now we have after applying augmentation that augmentation, we have 618 images in the training set, 30 00:02:03,000 --> 00:02:08,000 64 images in the validation set and 30 images the test set. 31 00:02:08,000 --> 00:02:08,000 Okay. 32 00:02:08,000 --> 00:02:12,000 So if we go to the health check as well, we can see that. 33 00:02:13,000 --> 00:02:18,000 So it here we don't have the issue of class balance because if we have multi classes, then we check 34 00:02:18,000 --> 00:02:20,000 whether the class is balanced or not. 35 00:02:20,000 --> 00:02:22,000 So here we don't have such issue. 36 00:02:22,000 --> 00:02:29,000 So we will export this dataset from Roboflow into our Google CoLab notebook to export the dataset, 37 00:02:29,000 --> 00:02:33,000 click on, download the dataset, and then select the model model. 38 00:02:33,000 --> 00:02:39,000 So as YOLO, V5 and YOLO V8, both were released by Ultralytics so we can select any of the model format 39 00:02:39,000 --> 00:02:47,000 here for implementing YOLO V8 to train the YOLO V8 or fine tune the V8 model on display dataset, then 40 00:02:47,000 --> 00:02:53,000 export the dataset in any format like you can select the V5 PyTorch format or YOLO V8 format. 41 00:02:53,000 --> 00:02:57,000 So I'll select the Yolov5 PyTorch format and click on Continue. 42 00:02:58,000 --> 00:03:04,000 Then in the next step, I will just copy this, all this code from here and paste into the Google CoLab 43 00:03:04,000 --> 00:03:04,000 notebook. 44 00:03:04,000 --> 00:03:12,000 But if you are doing in any other like I mean, if you are doing in PyCharm or if you are doing in Anaconda 45 00:03:12,000 --> 00:03:17,000 Navigator or if you are doing in or doing in Visual Studio, then you just need to copy this terminal 46 00:03:17,000 --> 00:03:17,000 command. 47 00:03:17,000 --> 00:03:23,000 But if you are doing in the Google CoLab notebook or Jupyter Notebook, just copy this from here and 48 00:03:23,000 --> 00:03:25,000 paste it in the Google CoLab notebook. 49 00:03:25,000 --> 00:03:27,000 So let's move towards the code part. 50 00:03:27,000 --> 00:03:34,000 So first of all, I will download the license plate automatic number, plate detection recognition required 51 00:03:34,000 --> 00:03:37,000 files from the Google Drive into my Google CoLab notebook. 52 00:03:37,000 --> 00:03:39,000 So these are the eight required files. 53 00:03:39,000 --> 00:03:45,000 It contain the plots, files or the V8 Pre-trained model files and the Ultralytics YOLO V8 required 54 00:03:45,000 --> 00:03:46,000 files over here. 55 00:03:46,000 --> 00:03:50,000 So you can see over here we have all these files currently using the zip format. 56 00:03:50,000 --> 00:03:56,000 So I will unzip all this file by just running this cell so it will unzip all the files and you can see 57 00:03:56,000 --> 00:03:59,000 over here we have all the files in the unzipped format. 58 00:03:59,000 --> 00:04:02,000 So it contains the file requirements file. 59 00:04:03,000 --> 00:04:03,000 Okay. 60 00:04:03,000 --> 00:04:08,000 So now I will import this required library from ipython dot display import in which. 61 00:04:08,000 --> 00:04:10,000 So why we need this libraries. 62 00:04:10,000 --> 00:04:14,000 If you want to display any input or output image into the Google colab notebook, then we need this 63 00:04:14,000 --> 00:04:16,000 import image library over here. 64 00:04:17,000 --> 00:04:17,000 Okay. 65 00:04:17,000 --> 00:04:19,000 So I have unzipped this folder. 66 00:04:19,000 --> 00:04:22,000 Now I need to set this folder as my current directory. 67 00:04:22,000 --> 00:04:24,000 So I'm just setting this folder as my current directory. 68 00:04:24,000 --> 00:04:30,000 Now in the next step I will install all the dependencies or you can say the requirements so it will 69 00:04:30,000 --> 00:04:34,000 install all the required libraries that are required to run this project successfully. 70 00:04:34,000 --> 00:04:39,000 If you don't install the dependencies, then when you run the training validation or the prediction 71 00:04:39,000 --> 00:04:44,000 step, then you will face the error that the following library is not installed or the Hydra library 72 00:04:44,000 --> 00:04:45,000 is not installed. 73 00:04:45,000 --> 00:04:47,000 Seaborn library is not installed. 74 00:04:47,000 --> 00:04:48,000 Matplotlib library is not installed. 75 00:04:48,000 --> 00:04:52,000 So it is important to install all the required libraries. 76 00:04:53,000 --> 00:04:56,000 So to install all the required libraries you just need to run the. 77 00:04:57,000 --> 00:05:03,000 A setup.py file, so it will install all the required libraries that are required to run this project. 78 00:05:03,000 --> 00:05:07,000 Some libraries are by default installed in the Google CoLab, but which are not installed. 79 00:05:07,000 --> 00:05:13,000 You can see here these are being installed, so if you miss this step, you can then you will face this 80 00:05:13,000 --> 00:05:16,000 error that the following library is not installed when you run the training script. 81 00:05:16,000 --> 00:05:17,000 So you need to install. 82 00:05:17,000 --> 00:05:19,000 Then you will need to install all the required libraries. 83 00:05:19,000 --> 00:05:25,000 So it's always better to run this step and install all the required libraries before running the script. 84 00:05:26,000 --> 00:05:34,000 Okay, so all the required libraries are being installed, so the dependencies are being installed. 85 00:05:34,000 --> 00:05:39,000 So now as we are doing detection over here, so I will go to Ultralytics, YOLO V8. 86 00:05:39,000 --> 00:05:45,000 So as we are doing detection, so we will go towards this folder and select set this folder as my current 87 00:05:45,000 --> 00:05:52,000 directory because it contains the training validation and prediction script and I am using as I, I'm 88 00:05:52,000 --> 00:05:53,000 doing detection over here. 89 00:05:53,000 --> 00:05:57,000 So I will just go to the detect folder which contains all my required scripts. 90 00:05:58,000 --> 00:05:58,000 Okay. 91 00:05:58,000 --> 00:06:01,000 So I will just check this folder as my current directory. 92 00:06:01,000 --> 00:06:07,000 So now I'm just exporting this dataset from roboflow the CoLab notebook. 93 00:06:07,000 --> 00:06:11,000 So just need to go here, click on copy and just paste this over here. 94 00:06:12,000 --> 00:06:12,000 Okay. 95 00:06:12,000 --> 00:06:18,000 And just if you run this, it will install the dataset or it will import the dataset from Roboflow into 96 00:06:18,000 --> 00:06:20,000 your Google CoLab notebook. 97 00:06:20,000 --> 00:06:24,000 So the dataset will be downloaded in this detect folder. 98 00:06:24,000 --> 00:06:29,000 The dataset will be downloaded in zip format, but it will automatically be unzipped when you just run 99 00:06:29,000 --> 00:06:30,000 this cell. 100 00:06:30,000 --> 00:06:31,000 So might take few seconds. 101 00:06:32,000 --> 00:06:38,000 So now you can see that the data set is being downloaded in the zip format, but it will be automatically 102 00:06:38,000 --> 00:06:39,000 unzipped here as well. 103 00:06:39,000 --> 00:06:41,000 So this might take few seconds. 104 00:06:41,000 --> 00:06:42,000 So let's see. 105 00:06:42,000 --> 00:06:45,000 As it completes, we will then proceed further. 106 00:06:45,000 --> 00:06:48,000 Okay, so dataset is downloaded in the zip format. 107 00:06:48,000 --> 00:06:54,000 So it will now automatically unzip over here and we will see this folder over here in the detect folder. 108 00:06:54,000 --> 00:06:57,000 So you can see the car license plate dataset over here. 109 00:06:57,000 --> 00:06:59,000 It's almost done. 110 00:06:59,000 --> 00:07:01,000 We have the train test validation folder. 111 00:07:01,000 --> 00:07:04,000 Each folder contain the images as well as the label. 112 00:07:04,000 --> 00:07:08,000 And then we have a data dot file which contains the class name. 113 00:07:08,000 --> 00:07:13,000 We have only one class, which is license, and here we have the train validation test, which is path, 114 00:07:13,000 --> 00:07:15,000 and this is the link of the dataset. 115 00:07:15,000 --> 00:07:18,000 So if you open this link, the dataset will open. 116 00:07:18,000 --> 00:07:19,000 Okay. 117 00:07:21,000 --> 00:07:27,000 Okay, so it's about to download the dataset and then we will go towards the training and then we will 118 00:07:27,000 --> 00:07:31,000 train our model on this dataset. 119 00:07:31,000 --> 00:07:33,000 So this might take few more seconds. 120 00:07:33,000 --> 00:07:34,000 So I'm just waiting. 121 00:07:34,000 --> 00:07:37,000 So as it gets completes, then I will proceed further. 122 00:07:38,000 --> 00:07:39,000 Okay. 123 00:07:39,000 --> 00:07:41,000 So it's about to download, so. 124 00:07:43,000 --> 00:07:45,000 So let me set the. 125 00:07:45,000 --> 00:07:48,000 Okay, so just run these cells as well. 126 00:07:49,000 --> 00:07:50,000 Okay. 127 00:07:53,000 --> 00:07:57,000 So now basically we need to in the next step, the dataset is downloaded. 128 00:07:57,000 --> 00:08:04,000 Now we need to train our model on this dataset, which is called a sense two dataset over here. 129 00:08:04,000 --> 00:08:08,000 So this is the Python Train.py script and we are using the V8 model. 130 00:08:08,000 --> 00:08:11,000 So there are five different V8 model. 131 00:08:11,000 --> 00:08:16,000 So the V8 end is the smallest, but it is less accurate and it is fast, although. 132 00:08:16,000 --> 00:08:21,000 But YOLO v8 X is the more most accurate and it is less fast. 133 00:08:21,000 --> 00:08:24,000 So we are using a V8 medium model. 134 00:08:24,000 --> 00:08:24,000 Okay. 135 00:08:24,000 --> 00:08:29,000 And we are setting the default image size because if we change the image size then we might compromise 136 00:08:29,000 --> 00:08:32,000 on accuracy and we are training our V8 model. 137 00:08:32,000 --> 00:08:38,000 Or you can say we are fine tuning the YOLO V8 model on license plate dataset and we are training the 138 00:08:39,000 --> 00:08:43,000 YOLO V8 model on license plate dataset for 120 epochs. 139 00:08:43,000 --> 00:08:49,000 Okay, so let's get started and let's train the V8 model on the license plate dataset. 140 00:08:49,000 --> 00:08:51,000 So it's about to start over here. 141 00:08:53,000 --> 00:08:57,000 First of all, it will download the model from official repository. 142 00:08:57,000 --> 00:09:01,000 It's downloading currently, so it might take some time. 143 00:09:01,000 --> 00:09:01,000 Okay. 144 00:09:01,000 --> 00:09:06,000 Then in the next step the training will start and the training will proceed a step by step. 145 00:09:06,000 --> 00:09:12,000 It will first work for one epoch and then for the second epoch. 146 00:09:12,000 --> 00:09:15,000 The training will take around a around two hours. 147 00:09:15,000 --> 00:09:19,000 So as the training start, I will pause the video and when the training completes, I will be back and 148 00:09:19,000 --> 00:09:21,000 I will explain you the rest of the code. 149 00:09:21,000 --> 00:09:23,000 So the training has started. 150 00:09:23,000 --> 00:09:24,000 It will take time. 151 00:09:24,000 --> 00:09:28,000 So I'm going to pause the video and as the training completes, I will be back. 152 00:09:31,000 --> 00:09:37,000 So guys, the training of the Yolo v eight model on the license plate dataset is completed and here 153 00:09:37,000 --> 00:09:38,000 are the results. 154 00:09:38,000 --> 00:09:44,000 We have trained our model on 120 epochs, which explained a year and we have selected the image size 155 00:09:44,000 --> 00:09:46,000 as 640 as YOLO. 156 00:09:46,000 --> 00:09:53,000 V eight model is officially trained on the image size of 640 by 640, so it's always better to set the 157 00:09:53,000 --> 00:09:58,000 image size as 640 when you fine tune your YOLO V8 model on any custom data set. 158 00:09:58,000 --> 00:10:04,000 So here we are fine tuning our YOLO V8 model on the license plate data set, and we have trained the 159 00:10:04,000 --> 00:10:07,000 YOLO V8 model on the license plate data set. 160 00:10:07,000 --> 00:10:08,000 And here are the results. 161 00:10:08,000 --> 00:10:15,000 We have trained our V8 model on 120 epochs and the mean average precision which we obtained is zero 162 00:10:15,000 --> 00:10:18,000 point with mean average precision. 163 00:10:18,000 --> 00:10:30,000 With IOU, 50 is 0.916, which is 91.6% and mean average precision when varies from 50 to 95 is 0.564, 164 00:10:30,000 --> 00:10:32,000 which is 56.4%. 165 00:10:32,000 --> 00:10:36,000 So the mean average precision is fine and the weights are saved. 166 00:10:36,000 --> 00:10:42,000 Over here, if you go to runs, detect, train and here we can see the weights file. 167 00:10:42,000 --> 00:10:44,000 Here we have the best and last weights. 168 00:10:44,000 --> 00:10:49,000 So the best.is updates file on which our model gives us the best results. 169 00:10:49,000 --> 00:10:55,000 And the last output is the weights file on the last epoch on the 1/20 epoch. 170 00:10:55,000 --> 00:11:02,000 So we will use best dot weights file for further validation of our trained YOLO V8 less model on the 171 00:11:02,000 --> 00:11:03,000 license plate dataset. 172 00:11:03,000 --> 00:11:05,000 So here is the model summary. 173 00:11:05,000 --> 00:11:12,000 The mean average precision of the 92.6% with 50 is obtained and mean average precision. 174 00:11:12,000 --> 00:11:17,000 When IOU varies from 50 to 95 is 57.6%. 175 00:11:17,000 --> 00:11:18,000 So the results are what? 176 00:11:19,000 --> 00:11:25,000 So all our after training, all the confusion matrix validation predictions, training, batch predictions, 177 00:11:25,000 --> 00:11:30,000 precision recall curve, F1 curves are saved in this folder. 178 00:11:30,000 --> 00:11:33,000 The train folder where you can see all these files. 179 00:11:33,000 --> 00:11:38,000 So let me check all these files in that train folder and display all these files in the Google CoLab 180 00:11:38,000 --> 00:11:39,000 notebook. 181 00:11:39,000 --> 00:11:45,000 So just if you click over here so it will display all the files which we have in the train folder. 182 00:11:45,000 --> 00:11:50,000 We have the confusion matrix, we have the F1 curve, we have the precision recall curve, we have the 183 00:11:50,000 --> 00:11:52,000 AACR, we have the results dot CSV file. 184 00:11:52,000 --> 00:11:59,000 So if I show you the result dot csv file, it contains the model evaluation or the model performance 185 00:11:59,000 --> 00:12:01,000 evaluation on each epoch after each epoch. 186 00:12:01,000 --> 00:12:05,000 So you can say after zero first, two, three, nine. 187 00:12:05,000 --> 00:12:09,000 So if you go over here, we have ten pages of this CSV file. 188 00:12:09,000 --> 00:12:16,000 So if I go to the last page, we have the on the last epoch 119 epoch, we have the box loss, the loss 189 00:12:16,000 --> 00:12:19,000 DF loss, precision recall. 190 00:12:19,000 --> 00:12:25,000 So we have all these values mean average precision with 50 to 95 mean average precision with 50. 191 00:12:25,000 --> 00:12:33,000 So we have all these basically in the CSV, we on the model performance after each epoch is saved. 192 00:12:33,000 --> 00:12:37,000 Okay, the results dot csv, the model performance after each epoch is saved. 193 00:12:37,000 --> 00:12:40,000 And then we have the confusion matrix over here. 194 00:12:41,000 --> 00:12:46,000 I have just displayed this confusion matrix over here into the Google CoLab notebook over here. 195 00:12:46,000 --> 00:12:52,000 So Confusion Matrix basically tells us how or how our model handles different classes. 196 00:12:52,000 --> 00:12:55,000 So in this case, we have only one class, which is lessons. 197 00:12:55,000 --> 00:13:02,000 So we can see that 84% of the time our model detected correctly that there is a lesson split, while 198 00:13:02,000 --> 00:13:08,000 16% of the times when there was a lesson played, our model was unable to identify it. 199 00:13:08,000 --> 00:13:14,000 So 84% of times our model was trained YOLO model detected correctly that there is a license plate, 200 00:13:14,000 --> 00:13:21,000 while 16% of the time when the license plate was there, our model was unable to detect it. 201 00:13:22,000 --> 00:13:28,000 So here we have the results dot PNG, which can you can see over here I've just displayed in the Google 202 00:13:28,000 --> 00:13:34,000 CoLab notebook so you can see that the Los loss box logs are decreasing, okay? 203 00:13:34,000 --> 00:13:37,000 They are decreasing and the mean average precision is on the rise. 204 00:13:37,000 --> 00:13:44,000 Mean average precision of 50 mean average precision when the varies from 50 to 95 is on the rise. 205 00:13:44,000 --> 00:13:50,000 So it means that if we train our model on 250 or 300 epochs, the mean average precision will further 206 00:13:50,000 --> 00:13:51,000 improve. 207 00:13:51,000 --> 00:13:55,000 Okay, so here I have the model prediction on the validation batch. 208 00:13:57,000 --> 00:14:02,000 So, uh, so basically these are the model predictions on the validation batch images. 209 00:14:02,000 --> 00:14:08,000 So these images are not used for the training, So it's always better to have a look and see how our 210 00:14:08,000 --> 00:14:11,000 model performs on the validation dataset, batch images. 211 00:14:11,000 --> 00:14:12,000 Okay. 212 00:14:12,000 --> 00:14:16,000 So here are some output examples from the validation batch images. 213 00:14:16,000 --> 00:14:21,000 So you can see that our model is able to detect successfully when there is a number plate and the confidence 214 00:14:21,000 --> 00:14:22,000 is very good as well. 215 00:14:22,000 --> 00:14:26,000 The confidence is most cases is above 0.9 and 0.8. 216 00:14:26,000 --> 00:14:27,000 Okay. 217 00:14:27,000 --> 00:14:32,000 So I have just added this section to just to save your projects into the Google Drive. 218 00:14:32,000 --> 00:14:35,000 And if you want to run this script after some time. 219 00:14:35,000 --> 00:14:39,000 So instead of training the model again, you can directly download the weights from the Google Drive 220 00:14:39,000 --> 00:14:40,000 and test your model. 221 00:14:40,000 --> 00:14:45,000 So I will just place my weights into the Google Drive and I have added the link over here. 222 00:14:45,000 --> 00:14:50,000 So if you want to test this, run the script on my weights, which are this batch stopped. 223 00:14:50,000 --> 00:14:56,000 So you can just download those weights by just running this cell and you can just test the model on 224 00:14:56,000 --> 00:14:56,000 this weights. 225 00:14:56,000 --> 00:14:57,000 Okay. 226 00:14:58,000 --> 00:15:01,000 So I'm just validating the custom model over here. 227 00:15:01,000 --> 00:15:06,000 So on the validation batch images, so on the test batch images, sorry. 228 00:15:06,000 --> 00:15:09,000 So basically these images are not used for the training. 229 00:15:09,000 --> 00:15:09,000 Okay. 230 00:15:09,000 --> 00:15:14,000 So it's always better to have a look and see how our results look like. 231 00:15:14,000 --> 00:15:17,000 So here we are validating our model on the validation batch images. 232 00:15:17,000 --> 00:15:20,000 So here are the validation batch basis. 233 00:15:20,000 --> 00:15:23,000 And on these images we are validating our model. 234 00:15:23,000 --> 00:15:28,000 Okay, so we are not using these images for that training and it's always better to see how our model 235 00:15:28,000 --> 00:15:30,000 performs on these images. 236 00:15:30,000 --> 00:15:30,000 Okay. 237 00:15:30,000 --> 00:15:34,000 So now what we have done is that we have. 238 00:15:36,000 --> 00:15:36,000 Okay. 239 00:15:36,000 --> 00:15:37,000 Let me first. 240 00:15:37,000 --> 00:15:38,000 So. 241 00:15:39,000 --> 00:15:40,000 That. 242 00:15:40,000 --> 00:15:40,000 Give me a minute. 243 00:15:40,000 --> 00:15:44,000 So here I'm just installing PIP install easy OCR. 244 00:15:44,000 --> 00:15:50,000 I'm using PIP install OCR because currently we have detected that there is a license plate like you 245 00:15:50,000 --> 00:15:55,000 can see that we have detected that this is a license plate, but we need to read the read the text inside 246 00:15:55,000 --> 00:15:56,000 the number plate. 247 00:15:56,000 --> 00:15:57,000 Okay. 248 00:15:57,000 --> 00:15:57,000 License plate. 249 00:15:57,000 --> 00:15:58,000 Okay. 250 00:15:58,000 --> 00:16:03,000 Basically we, we want to read the number plate like we want to read WOR516. 251 00:16:03,000 --> 00:16:03,000 Okay. 252 00:16:03,000 --> 00:16:10,000 So to read the text inside the number plate or the text inside the bounding box, which is over here. 253 00:16:10,000 --> 00:16:13,000 So to read the text of the license plate. 254 00:16:13,000 --> 00:16:17,000 Okay, so we use easy OCR to read the text inside the number plate. 255 00:16:18,000 --> 00:16:19,000 There are also other methods as well. 256 00:16:19,000 --> 00:16:26,000 You can use Tesseract OCR as well, but easy OCR gives best results, so we will use easy OCR to read 257 00:16:26,000 --> 00:16:29,000 the text of the number plate or of the license plate. 258 00:16:29,000 --> 00:16:29,000 Okay. 259 00:16:29,000 --> 00:16:33,000 As you can see that this project is license plate detection and recognition. 260 00:16:33,000 --> 00:16:38,000 So we need to detect the license plate and read the text of the license plate as well. 261 00:16:38,000 --> 00:16:41,000 So to read the text I will use first PIP install OCR. 262 00:16:43,000 --> 00:16:45,000 Okay, so it's already installed. 263 00:16:45,000 --> 00:16:46,000 I've run this. 264 00:16:46,000 --> 00:16:51,000 Okay, so here I'm just downloading a demo video in which I want to test my train model on the license 265 00:16:51,000 --> 00:16:55,000 plate dataset and here I have the predict.py script. 266 00:16:55,000 --> 00:16:57,000 So just run this script. 267 00:16:58,000 --> 00:17:00,000 So it might take a few seconds for the execution. 268 00:17:02,000 --> 00:17:05,000 Okay, so let's get started. 269 00:17:05,000 --> 00:17:09,000 So it's about to start and we have the best weights file over here. 270 00:17:09,000 --> 00:17:14,000 We are using the best weights file over here so you can see that the model is able to detect correctly 271 00:17:14,000 --> 00:17:17,000 the license plate, which you can see over here in each frame. 272 00:17:17,000 --> 00:17:23,000 The model is able to detect correctly that there is a license plate and the model has divided the whole 273 00:17:23,000 --> 00:17:28,000 video into 631 frames and currently it's on six 373 frames. 274 00:17:28,000 --> 00:17:34,000 So in this in this when we run this script, predict dot pi script, I'm basically detecting only the 275 00:17:34,000 --> 00:17:35,000 license plate. 276 00:17:35,000 --> 00:17:37,000 I'm not reading the text on the license plate. 277 00:17:37,000 --> 00:17:37,000 Okay. 278 00:17:38,000 --> 00:17:41,000 To read the text on the license plate we will do in the next phase. 279 00:17:41,000 --> 00:17:44,000 So currently I am only detecting the license plate. 280 00:17:44,000 --> 00:17:44,000 Okay. 281 00:17:44,000 --> 00:17:47,000 So here I will just display this demo video over here. 282 00:17:47,000 --> 00:17:52,000 So I have already staged this demo video over here if I run this script again. 283 00:17:52,000 --> 00:17:53,000 So it will take some time. 284 00:17:53,000 --> 00:17:58,000 So let me download this video and see is our model is able to detect the license plate or not. 285 00:17:58,000 --> 00:18:03,000 So just let me play this video and just navigate my screen towards this video. 286 00:18:03,000 --> 00:18:04,000 Okay. 287 00:18:04,000 --> 00:18:06,000 So now you can see the output video. 288 00:18:06,000 --> 00:18:10,000 So you can see that the model is able to check correctly that there is a license plate. 289 00:18:10,000 --> 00:18:15,000 Like, okay, you can see the bounding box and the model is able to read correctly that there is a license 290 00:18:15,000 --> 00:18:15,000 plate. 291 00:18:15,000 --> 00:18:18,000 So in this case, the model is giving us good result. 292 00:18:18,000 --> 00:18:22,000 But our main target is to read the text inside the license plate. 293 00:18:22,000 --> 00:18:25,000 Our main target is to read the text on the license plate. 294 00:18:25,000 --> 00:18:29,000 So let's move towards this and see how we can do this. 295 00:18:30,000 --> 00:18:37,000 So for this I will download the updated predict.py file which I have incorporated the code so I will 296 00:18:37,000 --> 00:18:40,000 download the updated predict.py file from the drive. 297 00:18:40,000 --> 00:18:42,000 So if I click over here. 298 00:18:42,000 --> 00:18:45,000 So this is the updated predict.py file. 299 00:18:47,000 --> 00:18:47,000 Okay. 300 00:18:47,000 --> 00:18:49,000 So if you want to. 301 00:18:49,000 --> 00:18:53,000 Okay, so just let me this line present result over here. 302 00:18:53,000 --> 00:18:54,000 Okay. 303 00:18:55,000 --> 00:18:57,000 You just click on Save now. 304 00:18:57,000 --> 00:19:02,000 Okay, so here you can see that we have created a function by the name of OCR image, and here we are 305 00:19:02,000 --> 00:19:08,000 initializing the OCR and we want to read the text in English, which is and we are using the GPU is 306 00:19:08,000 --> 00:19:14,000 equal to two because we are using GPU and easy OCR reader in means because we are reading the text which 307 00:19:14,000 --> 00:19:15,000 is in English. 308 00:19:15,000 --> 00:19:15,000 Okay. 309 00:19:15,000 --> 00:19:18,000 And here I've created a function OCR image. 310 00:19:18,000 --> 00:19:22,000 Okay, so here we are just choosing the coordinates of the bounding box. 311 00:19:22,000 --> 00:19:27,000 So if let me just open the paint and show you what we are basically doing over here. 312 00:19:30,000 --> 00:19:33,000 Okay, so just let me navigate my screen screen towards the pin. 313 00:19:33,000 --> 00:19:36,000 So if you go over here. 314 00:19:36,000 --> 00:19:41,000 Okay, so if, for example, this is my bounding box, okay, this is the bounding box. 315 00:19:41,000 --> 00:19:41,000 Okay? 316 00:19:41,000 --> 00:19:47,000 So in the bounding box, you these are my X1Y1 coordinates. 317 00:19:47,000 --> 00:19:52,000 These represent the X1Y1 coordinates, and these represent the X2Y2 coordinates. 318 00:19:52,000 --> 00:19:55,000 Okay, So in a bounding box, we have total four points. 319 00:19:55,000 --> 00:19:59,000 One is x, 1Y1, and this is the X2Y2 coordinates. 320 00:19:59,000 --> 00:19:59,000 Okay. 321 00:19:59,000 --> 00:20:05,000 So using these coordinates, we will update our code to read the text. 322 00:20:05,000 --> 00:20:06,000 Okay. 323 00:20:06,000 --> 00:20:08,000 So you can say that x over here. 324 00:20:09,000 --> 00:20:10,000 Let's write me over here. 325 00:20:11,000 --> 00:20:11,000 Okay. 326 00:20:11,000 --> 00:20:13,000 X here represents the x one. 327 00:20:14,000 --> 00:20:15,000 Okay. 328 00:20:15,000 --> 00:20:21,000 So just if I write arrow over here and y over here represents the Y one over here. 329 00:20:21,000 --> 00:20:26,000 So Y represents the Y one and W represents the x two over here. 330 00:20:26,000 --> 00:20:27,000 Okay. 331 00:20:27,000 --> 00:20:30,000 And x represents the y two over here. 332 00:20:30,000 --> 00:20:32,000 Okay, so this is how we have done. 333 00:20:32,000 --> 00:20:37,000 So these are the X1Y1 coordinates and this is the x2 and Y2 coordinates. 334 00:20:37,000 --> 00:20:42,000 Okay, so now what we are doing over here is that if you can see over here. 335 00:20:43,000 --> 00:20:45,000 If I go over here. 336 00:20:45,000 --> 00:20:50,000 Okay, So now from this complete view, reframe Now if you see over here, this is the complete frame. 337 00:20:50,000 --> 00:20:53,000 I'm just selecting the area where we have the bounding box. 338 00:20:53,000 --> 00:20:54,000 Okay? 339 00:20:54,000 --> 00:20:56,000 So if I explain this to you. 340 00:20:57,000 --> 00:21:00,000 Okay, So, okay, just let me go over here. 341 00:21:00,000 --> 00:21:02,000 So now you can see the video. 342 00:21:02,000 --> 00:21:05,000 So from this complete frame, this is the complete frame. 343 00:21:05,000 --> 00:21:08,000 I'm just selecting the area where I have the this bounding box. 344 00:21:08,000 --> 00:21:08,000 Okay. 345 00:21:08,000 --> 00:21:12,000 So I'm just selecting the area where I have this bounding box. 346 00:21:12,000 --> 00:21:12,000 Okay. 347 00:21:12,000 --> 00:21:13,000 Over here. 348 00:21:13,000 --> 00:21:19,000 So in this point, I'm just selecting the area where I have the bounding box where I have the license 349 00:21:19,000 --> 00:21:20,000 plate detected. 350 00:21:20,000 --> 00:21:20,000 Okay. 351 00:21:20,000 --> 00:21:23,000 Then I am converting this area into grayscale. 352 00:21:23,000 --> 00:21:29,000 Okay, so here I'm just converting this a area where I have the bounding box into grayscale and then 353 00:21:29,000 --> 00:21:31,000 I'm using easy OCR. 354 00:21:31,000 --> 00:21:36,000 I am just reading the text in the bounding box using easy OCR. 355 00:21:36,000 --> 00:21:39,000 I am just reading the text in this grayscale image of the bounding box. 356 00:21:39,000 --> 00:21:41,000 Okay, so. 357 00:21:42,000 --> 00:21:49,000 What I am doing is using first have only detected this area where I have this license plate. 358 00:21:49,000 --> 00:21:54,000 Then I have converted this area into grayscale and then I'm reading the text over here. 359 00:21:54,000 --> 00:21:56,000 What I have the text over here. 360 00:21:56,000 --> 00:22:01,000 You can see I'm just reading this text now using OCR, which I have inside this bounding box. 361 00:22:01,000 --> 00:22:02,000 Okay. 362 00:22:02,000 --> 00:22:08,000 So if I go back over here, so I'm just reading the text inside the bounding box which I have over here, 363 00:22:08,000 --> 00:22:09,000 which is over. 364 00:22:09,000 --> 00:22:12,000 You can see that n six, five, six is the text. 365 00:22:12,000 --> 00:22:13,000 Okay, so. 366 00:22:15,000 --> 00:22:19,000 Now, next up, what I'm doing is that I have already run this script so you can see that I have this 367 00:22:19,000 --> 00:22:20,000 printed. 368 00:22:20,000 --> 00:22:20,000 The result over here. 369 00:22:20,000 --> 00:22:22,000 These are the print results. 370 00:22:22,000 --> 00:22:22,000 Okay. 371 00:22:22,000 --> 00:22:26,000 I have already done this script before, so these are the print results. 372 00:22:26,000 --> 00:22:31,000 Okay, so what I'm doing is that you can see that, uh, this is the result. 373 00:22:31,000 --> 00:22:31,000 Okay. 374 00:22:31,000 --> 00:22:33,000 I have printed the result over here. 375 00:22:33,000 --> 00:22:35,000 This is the result after each of the frame. 376 00:22:35,000 --> 00:22:41,000 Okay, So what I'm doing is that so you can see that here I have the list over here. 377 00:22:42,000 --> 00:22:46,000 So if you can see over here, let me just a bit more. 378 00:22:46,000 --> 00:22:48,000 Let me just move it a bit more. 379 00:22:48,000 --> 00:22:48,000 Okay. 380 00:22:48,000 --> 00:22:50,000 So you can see that here. 381 00:22:50,000 --> 00:22:51,000 I have a complete list over here. 382 00:22:51,000 --> 00:22:52,000 Okay. 383 00:22:52,000 --> 00:22:57,000 So in the list, if we see over here, this represents the sub part of the list. 384 00:22:57,000 --> 00:22:57,000 Okay. 385 00:22:57,000 --> 00:23:02,000 So this is one this is the first element of the list, like zero element. 386 00:23:02,000 --> 00:23:04,000 This is the one element and this is the two element. 387 00:23:04,000 --> 00:23:09,000 So this is basically the confidence and this is basically the zero element of the list. 388 00:23:09,000 --> 00:23:09,000 Okay? 389 00:23:09,000 --> 00:23:10,000 This is the zero element. 390 00:23:10,000 --> 00:23:14,000 This is the first element and this is the second element. 391 00:23:14,000 --> 00:23:14,000 Okay. 392 00:23:14,000 --> 00:23:19,000 So I am just going this basically in result we have the list result contains the list. 393 00:23:19,000 --> 00:23:25,000 So I am just going through each, uh, each, each element in the list one by one. 394 00:23:25,000 --> 00:23:25,000 Okay. 395 00:23:25,000 --> 00:23:30,000 So if I'm saying that if the length of the, uh, this result is equal to one. 396 00:23:30,000 --> 00:23:35,000 Okay, but you can see here the length of the this result list is equal to three because this is the 397 00:23:35,000 --> 00:23:36,000 first element of the list. 398 00:23:36,000 --> 00:23:41,000 If we copy this over here, this is the second element of the list, and this is the third element of 399 00:23:41,000 --> 00:23:41,000 the list. 400 00:23:41,000 --> 00:23:46,000 But in that case, if the length of the list is one, then we will read the text in this. 401 00:23:46,000 --> 00:23:47,000 Okay. 402 00:23:47,000 --> 00:23:50,000 This is our basically we want to read this text of the list. 403 00:23:50,000 --> 00:23:51,000 Okay. 404 00:23:51,000 --> 00:23:56,000 But if the length of the result list is greater than one, so currently this length is equal to three 405 00:23:56,000 --> 00:23:57,000 or. 406 00:23:58,000 --> 00:24:00,000 And plus and and length of the result. 407 00:24:00,000 --> 00:24:03,000 One like this is a length of this text is greater than six. 408 00:24:03,000 --> 00:24:08,000 But here you can see that length is equal to four and and the confidence is greater than two. 409 00:24:08,000 --> 00:24:14,000 Then we want to display this text over here, so only we want to display that number plate text we want 410 00:24:14,000 --> 00:24:15,000 don't want this coordinates. 411 00:24:15,000 --> 00:24:20,000 So here we have this coordinates over here, but we don't want this coordinates. 412 00:24:20,000 --> 00:24:24,000 We only want to display that text on the number plate. 413 00:24:24,000 --> 00:24:27,000 We have basically what these coordinates represents. 414 00:24:27,000 --> 00:24:32,000 Okay, Let me tell you, uh, if you can see over here, if I create this bounding box over here, for 415 00:24:32,000 --> 00:24:35,000 example, this is my bounding box, So. 416 00:24:36,000 --> 00:24:37,000 If I just. 417 00:24:37,000 --> 00:24:38,000 Right over here. 418 00:24:38,000 --> 00:24:39,000 Okay, just let me. 419 00:24:39,000 --> 00:24:39,000 Sorry. 420 00:24:39,000 --> 00:24:43,000 I've just will not go on to the paint section, so if you can see over here. 421 00:24:45,000 --> 00:24:50,000 Okay, so if I create a bounding box over here, for example, this is my bounding box over here. 422 00:24:50,000 --> 00:24:52,000 So this represents the one. 423 00:24:52,000 --> 00:24:55,000 Okay, This is the second element of the list. 424 00:24:55,000 --> 00:24:59,000 This is the third element of the list, and this is the fourth element of the list. 425 00:24:59,000 --> 00:25:01,000 So these are the one, two, three, four. 426 00:25:01,000 --> 00:25:02,000 This is one, basically. 427 00:25:02,000 --> 00:25:02,000 Okay. 428 00:25:03,000 --> 00:25:05,000 So if I go back to the code again. 429 00:25:06,000 --> 00:25:07,000 So basically. 430 00:25:09,000 --> 00:25:11,000 If you go if you can see over here. 431 00:25:12,000 --> 00:25:13,000 Okay. 432 00:25:13,000 --> 00:25:18,000 So just let me go back over here so if you can see that this is the first element of the have shown 433 00:25:18,000 --> 00:25:19,000 the bounding box. 434 00:25:19,000 --> 00:25:21,000 This is the second element of the bounding box. 435 00:25:21,000 --> 00:25:23,000 This is the third element of the bounding box. 436 00:25:23,000 --> 00:25:25,000 And this is the fourth element coordinates of the bounding box. 437 00:25:25,000 --> 00:25:27,000 So this is the fourth coordinate. 438 00:25:27,000 --> 00:25:30,000 These are the coordinate points of the bounding box which we have drawn. 439 00:25:30,000 --> 00:25:34,000 Okay, so these are the four corner points of the bounding box. 440 00:25:34,000 --> 00:25:35,000 Okay? 441 00:25:35,000 --> 00:25:39,000 So you can see that over here we have four corner points on the bounding box. 442 00:25:39,000 --> 00:25:40,000 This is the first corner point. 443 00:25:40,000 --> 00:25:42,000 This is the second corner point. 444 00:25:42,000 --> 00:25:45,000 This is the third corner point and this is the fourth corner points. 445 00:25:45,000 --> 00:25:50,000 Okay, So if you can see over here, these are the four corner points on the bounding box. 446 00:25:50,000 --> 00:25:54,000 And this is a text we will just only display this text and this is the confidence. 447 00:25:54,000 --> 00:25:55,000 Okay? 448 00:25:55,000 --> 00:25:58,000 So we will call this function, which we have created over here. 449 00:25:58,000 --> 00:25:59,000 Okay. 450 00:25:59,000 --> 00:26:03,000 So we are saying that we only want to display that in the label section. 451 00:26:03,000 --> 00:26:10,000 We want to display the text inside the bounding box only we want to read the text of the license plate. 452 00:26:10,000 --> 00:26:11,000 So this will be the label. 453 00:26:11,000 --> 00:26:15,000 So if we just display this demo video over here. 454 00:26:16,000 --> 00:26:21,000 So now you can see that our model is able to read the text inside the number plate over here. 455 00:26:21,000 --> 00:26:24,000 Okay, So let me download this video and show you the results as well. 456 00:26:30,000 --> 00:26:33,000 Okay, Just let me navigate my screen over here. 457 00:26:35,000 --> 00:26:38,000 Okay, now, now you can see those results, like we said. 458 00:26:38,000 --> 00:26:42,000 So now you can see that our model is able to read the text inside the number plate. 459 00:26:42,000 --> 00:26:47,000 Okay, So our model is successfully able to read the text inside the number plate. 460 00:26:47,000 --> 00:26:51,000 So it's working completely fine and the results are very good as well. 461 00:26:51,000 --> 00:26:56,000 So that's all from this video tutorial, we are able to detect the license plate. 462 00:26:56,000 --> 00:26:58,000 We have trained our model to detect the license plate. 463 00:26:58,000 --> 00:27:02,000 We have read the text inside the license plate as well using easy OCR. 464 00:27:03,000 --> 00:27:05,000 See you in the next video tutorial. 465 00:27:05,000 --> 00:27:06,000 Till then, bye bye.