YOLOv5 training with custom data

แชร์
ฝัง
  • เผยแพร่เมื่อ 24 พ.ย. 2024

ความคิดเห็น • 590

  • @pranavkul525
    @pranavkul525 2 ปีที่แล้ว +34

    Thanks a lot for this video! Not only did you present everything really well, but also there wasn't anything in this video that I missed or didn't understand. It's good to be learning such complicated implementations with such ease!

    • @umar_muhammad_yaree
      @umar_muhammad_yaree 2 ปีที่แล้ว +1

      Please do you know how i can display number of objects in the video

    • @javlontursunov6527
      @javlontursunov6527 2 ปีที่แล้ว

      @@umar_muhammad_yaree I think you should count number of objects in each frame

    • @umar_muhammad_yaree
      @umar_muhammad_yaree 2 ปีที่แล้ว

      Yes
      Please do you know how i can do that

    • @davisonyeoguzoro9232
      @davisonyeoguzoro9232 ปีที่แล้ว

      @@umar_muhammad_yaree I am interested in this too

  • @aniketborkar9185
    @aniketborkar9185 2 ปีที่แล้ว +13

    For the first time I could execute something in the very attempt! Thanks to you Sir!!! This was seriously awesome.

  • @BenceTGIF
    @BenceTGIF ปีที่แล้ว

    The official colab notebook has changed since this video was made but still very useful! Thank you for your work!

  • @chowdhuryalvikhan1666
    @chowdhuryalvikhan1666 2 ปีที่แล้ว +9

    You explained this quite well man. Thanks a lot for such a great video. I do not comment on any channel but I couldn't stop myself from commenting on this video of yours. You are doing great man, best of luck for the future. Please do make videos on object detection using Reinforcement learning, it would help me a lot. I subscribed to your channel and liked your video, thanks man.

  • @sowmyar7056
    @sowmyar7056 2 ปีที่แล้ว +5

    Thank you for such a amazing video about YOLO, had almost everything needed in it ❤

  • @blackdiamond7790
    @blackdiamond7790 5 หลายเดือนก่อน

    Thank you soooo much. Finally after two days, I was able to do my class project ❤

  • @GaryWee111
    @GaryWee111 2 ปีที่แล้ว +2

    Wow, this explaination is one of the best I can find on TH-cam! Thanks for your effort!

  • @shapeyrthoughtsllshapeyrli1767
    @shapeyrthoughtsllshapeyrli1767 ปีที่แล้ว

    Thank you so much bro. I can't believe that I have done an object detection model in just one day.

  • @randomly-genrated
    @randomly-genrated 3 ปีที่แล้ว +38

    lol.....I got that memory error too, but I paused the video and never saw that you got it too and I spent an hour figuring out the same fix. Then I came back and resumed the video and in probably 10 seconds you changed the batch size lol....

  • @MarkMifsud
    @MarkMifsud 2 ปีที่แล้ว +1

    Thank you for the great tutorial and extra points for the chickens. I LOVE chickens!

  • @traeht
    @traeht ปีที่แล้ว

    Thank you very much! You helped me a lot with your clear explanation. I spent my whole day trying to get YOLOv5 up and running. Then I found your guide, followed it and everything works now.

    • @abprojects7477
      @abprojects7477 ปีที่แล้ว

      dear can you plz tell me how he unzipped the file i wrote the same code but it says unzip: cannot find or open ../train_data.zip, ../train_data.zip.zip or ../train_data.zip.ZIP.

    • @Isaac-be8dk
      @Isaac-be8dk 11 หลายเดือนก่อน

      i encounter the same error ,and the problem maybe we compress all the files together. and it works well when i seperate this files to upload ,hope this can help you.@@abprojects7477

    • @nandininandu8375
      @nandininandu8375 10 หลายเดือนก่อน

      ​Yup I've also similar error is there any sol

    • @BikuHazarika-w6i
      @BikuHazarika-w6i 3 หลายเดือนก่อน

      ​@@abprojects7477did u get the solution

    • @BikuHazarika-w6i
      @BikuHazarika-w6i 3 หลายเดือนก่อน

      ​@@nandininandu8375did u get the solution

  • @jackycwwang
    @jackycwwang 2 ปีที่แล้ว +3

    Wow, the explanation is so clear and straightforward! Thank you sir!!!

  • @hippieee9661
    @hippieee9661 5 หลายเดือนก่อน

    BEST TUTORIAL I EVER FOLLOWED! THANK YOUUUU!!!

  • @darklord_ice
    @darklord_ice 9 วันที่ผ่านมา

    Great video! I'm new to this. So with "val" you basically do the same thing again, just to train the model better? Which means, you could also put "negative" images that the model should exclude? I ask because I want to recognize sprocket holes on a film strip that are completely visible only, and the "half" ones should be recognized as invalid.

  • @mithraravi4722
    @mithraravi4722 2 ปีที่แล้ว +1

    Thankyou for this video! How to turn on Webcam and do object detection in real-time...can please say what changes should be done?

  • @elvenkim
    @elvenkim 2 ปีที่แล้ว

    1thanks a lot!! was looking for a simple way to detect object and found this channel. Subscribed

  • @mauricioramirezsalamanca8620
    @mauricioramirezsalamanca8620 2 ปีที่แล้ว

    Gracias por tomarte el tiempo de hacer este vídeo, no sabes lo útil que me fue, te juro, no tienes ni idea. Te amo.

  • @nehan5317
    @nehan5317 ปีที่แล้ว

    I have no words to express my gratitude towards you. Thanks a ton.

  • @zemenuketema1734
    @zemenuketema1734 ปีที่แล้ว

    Wow best explanation really am so happy by watching this video and many students get some thing especial.

  • @Eng-Ahmed_Gamal
    @Eng-Ahmed_Gamal ปีที่แล้ว +1

    Thank you for this great video, all the best 👍🏻

  • @-500IQgaming
    @-500IQgaming 10 หลายเดือนก่อน

    bro, I did this more than 4 hours and repeated many mistakes but I finally succeeded, that's an achievement for me :).

  • @peterkipropkurui9075
    @peterkipropkurui9075 2 ปีที่แล้ว

    Thank you very much, been looking for this explanation. Many thanks

  • @Yawanawa_K
    @Yawanawa_K 3 ปีที่แล้ว +1

    Thank you! This is just what I was looking for!

  • @Ashrafabazeed
    @Ashrafabazeed ปีที่แล้ว

    great video and explaining. I did everything as you mentioned but the model can't detect any image from the video. although it does detect it in the runs. .

  • @AleksandarBundalo
    @AleksandarBundalo 10 หลายเดือนก่อน

    Hello good man.
    I quickly tested your video, I like how you explain and kudos to you for your knowledge.
    I took a picture of a table lamp 10 times, 3 pictures are for validation and I recorded a video for 10 seconds, with an iPhone 13 Pro. I didn't succeed in processing the video, the pictures are trained... I don't know if it's because I converted MOV to MP4...or if there's something else at play.

  • @shivamamrutkar8915
    @shivamamrutkar8915 2 ปีที่แล้ว +1

    Hello Instructor, thanks for this amazing tutorial, please give us that video link of chickens, please upload it somewhere and provide us with link.
    Help appreciated...

  • @hrishikeshss7950
    @hrishikeshss7950 3 ปีที่แล้ว

    Brilliant Video, how to fine tune the model for face recognition?
    1. I'm getting 2 boxes on the same face for face recognition, for classes where the data was not sufficient
    2. Some faces are just labelled wrong

  • @magnetizing
    @magnetizing ปีที่แล้ว

    My man, you absolutely murdered this explanation. I would add #objectdetection, etc. to the description to help youtube/other people find this video faster (I was riding the struggle bus hard until I found this video lol). Cannot thank you enough!!!

  • @bharathijuvva5984
    @bharathijuvva5984 ปีที่แล้ว +1

    hello sir please do the same video for yolov8 also using webcam I mean for real time detection also

  • @scooterboom4157
    @scooterboom4157 ปีที่แล้ว

    you are so good at teaching!

  • @ku5h_
    @ku5h_ 2 ปีที่แล้ว +1

    thank you for the video!

  • @muhammadfahadnasir1627
    @muhammadfahadnasir1627 2 ปีที่แล้ว

    I cannot Thank You enough for this.

  • @lawalabdulazeez2636
    @lawalabdulazeez2636 2 ปีที่แล้ว +1

    Hello, thank you so much for this beautiful tutorial. Please, could you also do a tutorial on how to deploy this model live?

  • @herberthipolito9941
    @herberthipolito9941 2 ปีที่แล้ว

    Thankss. This video was really helpful in my project 😃.

  • @rubs63
    @rubs63 3 ปีที่แล้ว +1

    It's brilliant.!! Thanks for the amazing video!! Really helpful!!

  • @BernardLawes
    @BernardLawes 2 ปีที่แล้ว

    Very well done. Easy to follow.. you’ve got a new subscriber

  • @shaunincohst2839
    @shaunincohst2839 2 ปีที่แล้ว +1

    This TUTORIAL IS 100% WORKING GOOD TUTORIAL

  • @vijayk2065
    @vijayk2065 ปีที่แล้ว

    thank you, finally, I am able to identify chickens.

  • @osama82405
    @osama82405 2 ปีที่แล้ว

    excellent man,thank you for this video.look forward more such videos.

  • @umar_muhammad_yaree
    @umar_muhammad_yaree 2 ปีที่แล้ว

    Wow
    Thank you very much for this tutorial it helps alot
    It's very clear and straightforward 🤗🤗

  • @marthagichuki9539
    @marthagichuki9539 2 ปีที่แล้ว

    This is wonderful easy to follow tutorial. How can we tell the validation loss please

  • @noshinsaiyaraahmad3688
    @noshinsaiyaraahmad3688 ปีที่แล้ว

    Thank you so much for this useful tutorial!

  • @aleksandrjefremov3816
    @aleksandrjefremov3816 2 ปีที่แล้ว

    Thank you so much. Very clear explanation!

  • @RaulJayYanson
    @RaulJayYanson ปีที่แล้ว +1

    I have followed all the steps in the video, but my training only has 66% accuracy and when I try to predict and their is no prediction/bounding boxes. The results image has no labels.
    Does anybody know why is it? thanks in advance

  • @seinkoudiakite9546
    @seinkoudiakite9546 2 ปีที่แล้ว

    wahoo I'm stupefy, very good explanation. Thanks to you, I more learned about this topic

  • @MikeNapoli735
    @MikeNapoli735 3 ปีที่แล้ว +1

    Thank you very much! Nice video!

  • @demnatiyoussef9286
    @demnatiyoussef9286 ปีที่แล้ว +1

    come on man... you forgot a chicken in tha cage🍌

  • @DrShroom.
    @DrShroom. ปีที่แล้ว

    Hi this video was really good i found out i really like machine learning and that i chose my middle school right also i love the way you say chicken

  • @priyanshuupadhyay4854
    @priyanshuupadhyay4854 ปีที่แล้ว

    thanks for the explaination , i wanted to know that how can i count the objects detected in the image ,
    I have trained my model with your tutorial now want to count the objects detected in the image how can i do it ?

  • @hellosushant9471
    @hellosushant9471 2 ปีที่แล้ว

    Great video. Subscribed. I have 2 questions: 1. All images in train val dataset should be of same size. if it is not what to do? also, testing image size should be same or not?

  • @카자뷰
    @카자뷰 ปีที่แล้ว

    Thank you. It's settled thanks to you.

  • @jackycwwang
    @jackycwwang 2 ปีที่แล้ว

    Can you please make a tutorial about integrating DeepSORT with YOLOv5 for multiple-object tracking? Thx!!!

  • @Nour-Elislem
    @Nour-Elislem 2 ปีที่แล้ว

    Thank you bro for this tutorial ,you're Amazing

  • @SohailKhan-tt5eh
    @SohailKhan-tt5eh 2 ปีที่แล้ว

    Thank you for this amazing tutorial....................

  • @yasirali8409
    @yasirali8409 2 ปีที่แล้ว

    What a great Explanation. very thanks it help me a lot in sport analysis matches.!)

  • @medinikb
    @medinikb 2 ปีที่แล้ว

    Thank you so much........for sharing your knowledge.............

  • @עטרהחדד
    @עטרהחדד 7 หลายเดือนก่อน

    Thank you very much for the video it really helps me,
    I have a small question: if my computer does not support cuda how can I solve the errors due to the layers: dnn and fft?
    I would really appreciate it if you could help me, I really need it

  • @souadelh8123
    @souadelh8123 ปีที่แล้ว

    Hi thank you for this video. I have a question , it just first time I will learn about this . after detection is that number showed with detection is it a feature of chicken? Is it its length?

  • @Rahul-kc4gf
    @Rahul-kc4gf 2 ปีที่แล้ว +2

    I need guidance for Final year project how I detect and recognize multiple faces from single image. thank you in advance

    • @SoftwaresAI
      @SoftwaresAI  2 ปีที่แล้ว

      hi, i can help you in this. Contact me on google chat / google hangout , at biplob00110011@gmail.com

  • @caneron6539
    @caneron6539 ปีที่แล้ว

    Thank you so much bro. God bless you.
    I love you ♥

  • @tirthrajgandhakwala6201
    @tirthrajgandhakwala6201 2 ปีที่แล้ว +2

    Hello. I am trying to follow your video but some things have changed on colab
    This is my yaml file.
    train: ../train_data/images/train # train images (relative to 'path') 128 images
    val: ../train_data/images/val # val images (relative to 'path') 128 images
    test: # test images (optional)
    # Classes
    names:
    0: pothole
    This is the error I get in training part.
    Transferred 739/745 items from yolov5x.pt
    AMP: checks passed ✅
    optimizer: SGD(lr=0.01) with parameter groups 123 weight(decay=0.0), 126 weight(decay=0.0005), 126 bias
    albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01), CLAHE(p=0.01, clip_limit=(1, 4.0), tile_grid_size=(8, 8))
    train: Scanning '/content/train_data/labels/train' images and labels...0 found, 500 missing, 0 empty, 0 corrupt: 100% 500/500 [00:00 0 or not augment, f'{prefix}No labels found in {cache_path}, can not start training. {HELP_URL}'
    AssertionError: train: No labels found in /content/train_data/labels/train.cache, can not start training. See github.com/ultralytics/yolov5/wiki/Train-Custom-Data
    Please help me out. I really want to get this done for course project. Thank you

    • @WisdomWhispers807
      @WisdomWhispers807 ปีที่แล้ว

      Did you find any solution for this. I'm also getting this error.

  • @jiangliubin3543
    @jiangliubin3543 3 ปีที่แล้ว +3

    Thanks for your video, really helpful! I am doing multiple objects detection, and want to show the result pictures without any object name and object probability on the bounding box, how shall I change for this? Appreciate if you can share.

    • @ppal84
      @ppal84 3 ปีที่แล้ว +1

      Try passing the command with "--hide-labels" and "--hide-conf"

  • @abidmaqbool1427
    @abidmaqbool1427 3 ปีที่แล้ว +5

    good tutorial, keep going on man!

  • @girish2555
    @girish2555 3 ปีที่แล้ว +2

    Really a lifesaver 👍

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      If you like to support my channel, here is link www.buymeacoffee.com/biplob004

  • @epuchumoses321
    @epuchumoses321 3 ปีที่แล้ว +3

    thank you very much for the tutorial it was very beneficial. though i have a question @DeepLearning. how can i extract the custom_weights and custom_config files for deployment in object detection on a raspberry pi.

    • @marcelofsouza1
      @marcelofsouza1 3 ปีที่แล้ว

      You can use your weight file (last.pt or best.pt) and load this trained models using pytorch. After loading you can use it combined with opencv.

    • @tributemaker
      @tributemaker 3 ปีที่แล้ว

      @@marcelofsouza1 how can I get best.pt file. I cannot find it..

  • @merondesta3895
    @merondesta3895 2 ปีที่แล้ว

    thankyou sir that was useful. i trained using my own dataset but no matter how long i trained it the m@p will not get improved from 0.34 something like that i have used same data for yolov4 but it performs well

  • @TheMrkillermole
    @TheMrkillermole 3 ปีที่แล้ว +2

    Thanks man, very helpful video.

  • @AK-ux9rv
    @AK-ux9rv หลายเดือนก่อน

    When you trained the model with the custom dataset, will it still be able to detect the existing COCO objects like person, bus, car etc? Or after training will it ONLY detect chicken, f yes, then what to do so that it detects COCO objects + chicken ?

  • @NamLe-fl4sz
    @NamLe-fl4sz 10 หลายเดือนก่อน

    From VietNam. Thanks

  • @emilefourie8544
    @emilefourie8544 ปีที่แล้ว

    Great video, thanks man!

  • @javokhirmusaev721
    @javokhirmusaev721 3 ปีที่แล้ว +1

    Thanks man, really good example! Can you make video or give source that can be trained on local computer with gpu on custom dataset?

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      If you have installed PyTorch on your local computer and have a GPU, you can train model exact same way as i did in here. If you need any help contact me, i can help you on this project.

    • @alpertemizel9892
      @alpertemizel9892 3 ปีที่แล้ว +1

      @@SoftwaresAI how can i contact with you

    • @Grayando1
      @Grayando1 2 ปีที่แล้ว

      @@SoftwaresAI need help with this too

  • @umar_muhammad_yaree
    @umar_muhammad_yaree 2 ปีที่แล้ว

    How can you display the number of chiken in the video
    Thanks for the video
    I have a task which is pedestrian detection and congestion measurement
    I need to build a model that can detect pedestrian and also be able to say if the frame is congested
    For example let's say in a frame greater than 20 people means congested

  • @owusukwesi6398
    @owusukwesi6398 3 ปีที่แล้ว

    Thnaks for a wonderful video...
    Im done training the model, but to use the file in my opencv project, i need it in .weight format.....
    where do I get the file??

  • @ahmedalsaati3408
    @ahmedalsaati3408 3 ปีที่แล้ว

    Thank you very much. Very useful. Can I build a front end app to use it for users?

  • @amirkarim9932
    @amirkarim9932 ปีที่แล้ว

    It was great👍 thank you , keep it up

  • @samdb8393
    @samdb8393 3 ปีที่แล้ว +2

    Hi every one. Thanks a lot for this short and sweet Tutorial .
    I have a question and it is about image size in YOLOv5. is it in 640 *640 ? and can we use another size? if so , HOW?
    Thank you in advance

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      Yes, you can use any size you want
      do you want to take live tuition from me? If interested then reply to this comment.

  • @dogukanklc6645
    @dogukanklc6645 ปีที่แล้ว

    Bro you are the best . You are the damnnn best

  • @kemaloguzhanyeyit9018
    @kemaloguzhanyeyit9018 ปีที่แล้ว

    Great video ! How can I train my dataset with YOLO v3-tiny by using this guide ?

    • @mustafaturk9180
      @mustafaturk9180 ปีที่แล้ว +1

      Same way he did it here, but its better for you to just train your model on your IDE.

  • @namangoyal8477
    @namangoyal8477 2 ปีที่แล้ว

    Excellent tutorial.

  • @christinedai2915
    @christinedai2915 11 หลายเดือนก่อน

    Why do you not need a "test" folder? Aren't we supposed to have train, val and test sets? I'm confused.

  • @طيفالحارثي-ط5ل
    @طيفالحارثي-ط5ل 7 หลายเดือนก่อน

    You are amazing thank u so much ❤❤❤❤❤❤❤

  • @amathpati9107
    @amathpati9107 3 ปีที่แล้ว +1

    Your algorithm is working very fine, it gives outstanding results. But I also want to count the detected objects('bearings'). How can I do the same please reply as soon as possible.

    • @cosmilitebar6772
      @cosmilitebar6772 3 ปีที่แล้ว +1

      you can get the results from detection and write them to a List. Then the len(yourlist) is the number of things detected

    • @suryasingh425
      @suryasingh425 2 ปีที่แล้ว

      @@cosmilitebar6772 can you explain a bit more?

  • @quangnhatnguyen4508
    @quangnhatnguyen4508 ปีที่แล้ว +1

    how to unzip train data, sir? i can't do it in step .

  • @elaineyang2721
    @elaineyang2721 ปีที่แล้ว

    thank you! I have a question, why in the beginning you have train and val, but later you get train and test?

  • @efgerfgergferg6749
    @efgerfgergferg6749 3 ปีที่แล้ว +1

    Any tips on increasing the confidence level of detected objects? Thanks

  • @parbirmondal3723
    @parbirmondal3723 3 ปีที่แล้ว +1

    Thank you Sir... It is really to the point and informative... Really too helpful

  • @nikhildesai2460
    @nikhildesai2460 3 ปีที่แล้ว +1

    Excellent Work Brother..!!

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      Thanks ✌️

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      Please support my channel here www.buymeacoffee.com/biplob004

  • @herlina9622
    @herlina9622 2 ปีที่แล้ว +1

    could you please help me? i tried to unzip train data few times, it always comes out like this: unzip: cannot find or open ../train_data.zip, ../train_data.zip.zip or ../train_data.zip.ZIP.
    what should i do?

  • @vishvajitviswanathan7970
    @vishvajitviswanathan7970 2 ปีที่แล้ว

    For the same, how to use for live detection. That is using our own camera instead of a video!?

  • @faceprofesor
    @faceprofesor 3 ปีที่แล้ว +1

    Thanks for the video. Is it possible to annotate semantically or with polygons? Thanks

  • @yashraval8844
    @yashraval8844 ปีที่แล้ว

    Hey, great explanation. I have a doubt that how can we use the same detection file to run on multiple sources at the same time like you used 1 video source file can we detect objects in different videos together by using just one detection file instead of separate detection files for each video

  • @mariomasseiro9090
    @mariomasseiro9090 3 ปีที่แล้ว

    one of the best videos i used for my project, really thanks

  • @iitguycrypto9946
    @iitguycrypto9946 2 ปีที่แล้ว

    Great Video!! Is it possible to do this for Thermal Images as well. I would like to detect object in thermal images. Please advise.

  • @CrazyFanaticMan
    @CrazyFanaticMan ปีที่แล้ว

    Can you do a new video but with Yolov8? Ultralytics just released Yolov8

  • @alpcnar5877
    @alpcnar5877 ปีที่แล้ว

    You have Train and validation but where is the test set ?

  • @leandroleandro6844
    @leandroleandro6844 3 ปีที่แล้ว

    Very useful, tks for sharing!

  • @amarsaish5931
    @amarsaish5931 ปีที่แล้ว

    we didn't used labels while training the data, how it detected without taking our labels. If we want to use this for any other custom data how is it possible

  • @supriyantoku
    @supriyantoku 3 ปีที่แล้ว

    thank you so much. Very good tutorials.

  • @ДмитрийБолховский
    @ДмитрийБолховский 3 ปีที่แล้ว +3

    Thank you very much for the helpful video.
    How to download the final file with the trained model to your computer so that you can use it locally?

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว +2

      You are welcome, you can find trained weight at location:
      yolov5/runs/train/exp/weights/last.pt
      you will find exact path of your saved model, printed at last line when training completes.

    • @ДмитрийБолховский
      @ДмитрийБолховский 3 ปีที่แล้ว

      @@SoftwaresAI sorry for the next stupid question. The model was downloaded with the extension .pl
      Earlier I opened models with h5 extension like this:
      detector = ObjectDetection ()
      detector.setModelTypeAsTinyYOLOv3 ()
      detector.setModelPath ('yolo-tiny.h5')
      detector.loadModel ()
      How to open last.pt file in python?

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      No problem,
      Here is solution:
      Watch my this video th-cam.com/video/Qs_kMvU1IIo/w-d-xo.html here simply replace yolov5s.pt with your runs/train/exp/weights/last.pt model.
      like this: --weights runs/train/exp/weights/last.pt instead of --weights yolov5s.pt

    • @SoftwaresAI
      @SoftwaresAI  3 ปีที่แล้ว

      @@ДмитрийБолховский why you need to save model in .h5 format ?
      its automatically saved in .pt format, simply use that on your local computer.

    • @ДмитрийБолховский
      @ДмитрийБолховский 3 ปีที่แล้ว

      @@SoftwaresAI is it possible to save the trained model in h5 format

  • @ridergroup719
    @ridergroup719 3 ปีที่แล้ว

    The prediction which you were telling at 16:01 , it doesn't pridict anything.. can you give me the solution. bonding box was not created in prediction.