Пікірлер
@Kaixiang77441 3 сағат бұрын
how did you know the candy weight? in YOLO
@aryagumilang4155 22 сағат бұрын
Hi EJ Technology, does this tutorial still work with the current version of Python (3.11.11) on Google Colab? Some libraries are no longer supported in version 3.11.11 python 3.11.11 doesn't support pyyaml 5.3.1 anymore
@baguspandu_1ia244 23 сағат бұрын
cuse me sir when ive upload the data.zip with the same format direct to the collab but it doesnt read any files, i try the gdrive too Created folder at /content/data/train/images. Created folder at /content/data/train/labels. Created folder at /content/data/validation/images. Created folder at /content/data/validation/labels. Number of image files: 0 Number of annotation files: 0 Images moving to train: 0 Images moving to validation: 0
@baguspandu_1ia244 23 сағат бұрын
nevermind i did it
@mickfone4821 Күн бұрын
Thank you very much, it's so cool.
@madebymonii17 2 күн бұрын
That git hub link is asking username and password what to do for that
@creativearts4917 3 күн бұрын
you r the person i was looking for from almost 2 months... i have watched almost all this type vids... have you did any comparison between yolo/ultralytics and tensorflow?
@EdjeElectronics 2 күн бұрын
Ultralytics YOLO models are way better than TensorFlow detection models. They're faster, more accurate, more modern, and easier to train. I get 7 FPS with a 640x640 YOLO11n model on my Raspberry Pi 5 vs 4 FPS with a TensorFlow SSD-Mobilenet model. Here are a couple links to look at: TFLite comparison: www.ejtech.io/learn/tflite-object-detection-model-comparison YOLO comparison: kzbin.info/www/bejne/lYiuhGd7brigoaM
@TaymourElErian 3 күн бұрын
Great video, can you share the link to the video on using the model with Coral ?
@paulokaedolima2739 4 күн бұрын
You helped so much in my Cardgame Tracker! Thanks!
@ShreyasSachinKshatriya 4 күн бұрын
hands down the best tutorial for TensorFlow Lite, I learnt a lot thanks
@fredwrinkle 4 күн бұрын
"There's a few limitations..." ,,,,fourth, it's cheating, illegal, and could lead to a prison sentence if you bring it into a casino and use it. It's useless for online casinos because most online casinos only penetrate half the deck. And if an online casino suspects you're using software, they will not pay you out. It's in their terms of service which you agree to before you play. Good luck fighting a casino in a foreign court trying to get paid out.
@fernando2074 5 күн бұрын
Thank you so much!!! Not only great tutorial but also inspirational spirit, so grateful for this😊
@mr.rizcarn4633 6 күн бұрын
Amazing tutorial <3 ! But may I ask sir, why my model detects other things (like person, watch, etc.) that isn't belong to my custom data? I hope for your response sir, thank you
@EdjeElectronics 5 күн бұрын
Weird that it's detecting other objects! The model will only detect classes listed in the data.yaml file (created at 11:53 in the video). Double-check your classes.txt and the data.yaml file to confirm there aren't extra classes listed.
@mr.rizcarn4633 22 сағат бұрын
Hi sir it's all good now, thank you. But my new problem is, it detects other objects as one of my classes. For example: It detects the class name "bottle" even though I present a TIN CAN (which is not on my class list)
@BadHits-c5i 6 күн бұрын
What if I don't have a gpu then which pytorch to install
@EdjeElectronics 5 күн бұрын
Then you can just use "pip install ultralytics" (which we do at 17:20 in the video). This installs the CPU version of PyTorch by default.
@artoftech4221 6 күн бұрын
i am using raspberry pi zero 2w. when i try to install the requirements, after sometime, it shows installing dependencies and stops there, not going further. Is there any solution ? Please help its urgent!
@milankarakas3005 8 күн бұрын
It is amazing how good it works on Jetson Nano. I ordered this one, but delivery time is between 4 and 7 weeks. Oh, well...
@milankarakas3005 8 күн бұрын
Uh, that one is 6 years ago. Since then, there are many changes. For example, CUDA 12.6 is currently installed on my computer. I can't change that because other libraries and programs using exactly that CUDA. If you find time, please make another video explaining this with new upgrade. Thanks.
@krupaltisgaonkar238 8 күн бұрын
Great Video! I really need this for my ML project. However, I already have created a dataset using LabelImg and created xml files for a Tensorflow format. Do you know how I can change it to a YOLO format? NEW SUBSCRIBER!!!!!!!!!
@EdjeElectronics 7 күн бұрын
Thanks, glad it's helpful! Yes, here's one way to convert TensorFlow data (which is in Pascal VOC format) to YOLO format. These commands will download a conversion script and then run it. The arguments for the script are 1) the path to the labels directory, 2) the path to the classes.txt labelmap file, and 3) a name for an output file. When you run the script, it will create a .txt file with data in YOLO format from every .xml file in the labels directory. curl -o convert_voc_yolo.py raw.githubusercontent.com/geaxgx/playing-card-detection/refs/heads/master/convert_voc_yolo.py python convert_voc_yolo.py path/to/labels/dir classes.txt converted.txt Good luck :| if that's complicated, Roboflow can convert data formats too.
@krupaltisgaonkar238 7 күн бұрын
@@EdjeElectronics I also have two directories: a train and a test folder, should I combine both of them into one and label it images and let google colab seperate it?
@milankarakas3005 8 күн бұрын
Ha, my idea is little bit different; to detect all my three cats. That is because sometimes neighbors cats want to enter home and steal food. So far, it is really bad detection and I am not using Yolo, but rather pytorch. Instead sending SMS, I will use just audible buzzer or something simple to notify me about my cats wanting in or out. Also, all three are 'silent cats'. :D Oh, yes. Winter time, door is closed and it is transparent, so no need for two cameras (inside and outside). But summer time is problematic; many neighbor cats want in, eat and leave. I want to prevent that by using loud buzzer for which my cats will get used to, but neighbor cats maybe not.
@milankarakas3005 8 күн бұрын
After detecting, can you use python script to make an action? I have idea to "ring the bell" when one of my cats want to enter. There is a camera watching the door. Until now, I should to watch all the time, but what if Yolo can detect my cat and notify me? If not my cat, ignore.
@EdjeElectronics 8 күн бұрын
@@milankarakas3005 haha I made a video about that exact project! Check out my Raspberry Pi Pet Detector. It’s very dated now and the code will look different with YOLO. But yes it is certainly capable of that! kzbin.info/www/bejne/nXjUh4GrjtSfbNE
@ya_yo 9 күн бұрын
thanks for the tutorial!! would be great to have a tutorial to use these models in a iphone 🫣
@pludwig2820 9 күн бұрын
Great videos and well documented tutorials! I would like to ask if there is any way to automatically transform your PascalVoc formatted custom images ( made for Tensorflow ) into Yolo related ones ?
@EdjeElectronics 7 күн бұрын
Thank you! Yes, here's one way to convert Pascal VOC formatted data to YOLO format. These commands will download a conversion script and then run it. The arguments for the script are 1) the path to the labels directory, 2) the path to the classes.txt labelmap file, and 3) a name for an output file. When you run the script, it will create a .txt file with data in YOLO format from every .xml file in the labels directory. curl -o convert_voc_yolo.py raw.githubusercontent.com/geaxgx/playing-card-detection/refs/heads/master/convert_voc_yolo.py python convert_voc_yolo.py path/to/labels/dir classes.txt converted.txt Good luck :| if that's complicated, Roboflow can convert data formats too.
@APie357 9 күн бұрын
hes alive
@Jdmr1701 9 күн бұрын
Thank you for posting this. It’s very apropos for me at the moment. I am trying to train a model to identify our dogs for a larger project. I am new at the AI stuff and not a coder, but this is the kind of information I have been searching for the last several days. Thank you!
@oludelehalleluyah6723 9 күн бұрын
Yooo... It as been so long... I have learnt so much from you... I learnt how to build object detection models with tensorflow and it has gone a long way in helping me... Good to have you back
@PixlPainter_ 9 күн бұрын
Thanks a lot for your effort! This video is exactly what I needed, great job explaining! Can't wait for the article on deploying on a Raspberry Pi 🎉
@EdjeElectronics 9 күн бұрын
Thanks for the kind words, and I'm glad you found it useful! I'm shooting to have the Pi video out before the end of January. Cheers!
@shashankvray9042 9 күн бұрын
Good video. Thanks for the comprehensive explanation. Can you start a whole new playlist on GenAI and LLM models with more in-depth explanation that includes explanations on fine tuning methods like PEFT as well
@kessbeats603 9 күн бұрын
good to see you. keep it up
@ilya6319 10 күн бұрын
Hey @EdjeElectronics, which camera did you try? What is the best?
@EdjeElectronics 9 күн бұрын
I like using the Logitech c920 camera. It's reliable and has good image quality for the cost. www.amazon.com/Logitech-C920x-Pro-HD-Webcam/dp/B085TFF7M1
@ilya6319 9 күн бұрын
@@EdjeElectronics Thank you, your research are awesome,
@syazilikhai524 10 күн бұрын
Why it could not detect address file that has been built. For step 7 or 8,it says that no gouth file found. And after every step has been done, there is no images
@hafi1915 11 күн бұрын
Nice work, can u make a video how we can run a yolov11 model in Raspberrypi and jetson nano for realtime object detection task (using camera module)
@EdjeElectronics 11 күн бұрын
@@hafi1915 Thanks, yes, I’m planning to release a video showing how to run YOLO11 on the Raspberry Pi soon. Stay tuned.
@kyleheppler2860 11 күн бұрын
Welcome back! Keep going!
@AffectiveApe 12 күн бұрын
Awesome to see you back mate! I used your tensorflow tutorial way back during my phd for a little side project identifying mice in security camera footage. Sadly, the project didn't quite pan out and I had to move on to another project and abandon the ML, but I am excited to see how things have progressed! who knows, I still have the video footage, maybe I need to take another stab at it now that I've graduated 😂
@EdjeElectronics 12 күн бұрын
Thanks!! It's waay easier to train detection models nowadays, so if you could figure out TensorFlow you can definitely figure out YOLO 😃.
@MakerMadness 12 күн бұрын
LOVE that you’re back!
@EdjeElectronics 12 күн бұрын
Thanks buddy! It's good to be back. Cheers.
@EdjeElectronics 12 күн бұрын
Stay tuned for a new video on training YOLO models - it will be out in a few days!
@vigneshkarapurkar9155 12 күн бұрын
May I know what os are you using as I tried using buster on rpi3 the numpy library it stucks there
@SunnY-nv2lm 17 күн бұрын
can we increase the fps / speed on raspberry pi if yes how ...is there any way to accelerate the ml models in raspberry pi either hardware like coral or software or code changes ..please do make a video on this
@anonymousblox 18 күн бұрын
TechJoyce Pls build me a PC
@KeithLovely 23 күн бұрын
Is the code public? I have an idea to take it up a level
@EdjeElectronics 22 күн бұрын
@@KeithLovely nope, im building on it for the casinos now. But you can explore geaxgx’s repository to see how the card generation works if you want to try it yourself.
@fredwrinkle 4 күн бұрын
Yes. The code is 6. Tell everyone.
@KeithLovely 23 күн бұрын
Any chance the softwares online or guithub. Inhabe annidea to take this nect level
@benji492 23 күн бұрын
Could you please shoot a new setup video for the Raspberry Pi 5 with Debian Bookworm? Also, Google has moved the models from the TensorFlow site to a Google page.
@EdjeElectronics 22 күн бұрын
@@benji492 thanks for the heads up! I’m working on a video showing how to run YOLO models on the Pi, it should be out in a month.
@technoandsincesoreyansh 25 күн бұрын
can you do this with dialouge writing
@supasol999 25 күн бұрын
Did you find a way to have it count multiple decks? Can it be used with meta glasses or AR glasses?
@alialfaytouri3786 27 күн бұрын
can you explain how to add TFLite_Detection_PostProcess for this model
@ah-lx9xi Ай бұрын
Thanks for sharing video. Can this be performed with cpp most videos on have noticed are with python. Can this code be made to announce sounds messages of detected objects and where about in picture the detected object is? Thanks.
@mehmetakifturhan9106 Ай бұрын
Amazing tutorial, I'm looking forward to the quantization video.
@PhatHuynh-f2y Ай бұрын
Dear Edje! Thank you for your deep instruction! I got the .tflite from my custom dataset already. But when run the detection_webcam.py it always show the error: "/home/pi/tflite1/TFLite_detection_webcam.py", line 180, in <module> frame = frame1.copy()^^^^^^^^AttributeError: 'NoneType' object has o attibute 'copy' I failed on both coco dataset and my custom dataset. I suspect that the problem come from my Raspbrry Pi Camera Module V3 since your source look like match with webcam USB camera... Could you please advice any for my case? Thank agian for your strong support anh hope to get your advice soon! Thank Edje!
@MohammedModhawwi Ай бұрын
In step 5 the code kept running for almost 30 minuets without any improvement with the last five lines as this W1205 16:40:36.064376 132839187022464 deprecation.py:337] From /usr/local/lib/python3.10/dist-packages/tensorflow/python/util/dispatch.py:1082: sample_distorted_bounding_box (from tensorflow.python.ops.image_ops_impl) is deprecated and will be removed in a future version. Instructions for updating: `seed2` arg is deprecated.Use sample_distorted_bounding_box_v2 instead. WARNING:tensorflow:From /usr/local/lib/python3.10/dist-packages/tensorflow/python/util/dispatch.py:1082: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version. Instructions for updating: Use `tf.cast` instead. W1205 16:40:38.264565 132839187022464 deprecation.py:337] From /usr/local/lib/python3.10/dist-packages/tensorflow/python/util/dispatch.py:1082: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version. Instructions for updating: Use `tf.cast` instead. what is the problem?
@Minisatheesh Ай бұрын
Hi I am a high schooler and working on a project using Raspberry Pi 5 for object detection. I have trained a custom TFLite model using your Colab notebook. However, when running the detect_webcam.py file, I get the error "VideoType has no object Frame". For this, I've tried to utilize camera pipes in both h264 and mJPEG formats, but to no avail. Can you please help me, as I've tried so many options, but feel stuck? I am using a Raspberry Pi Camera Module 3.
@EdjeElectronics Ай бұрын
Hello, usually when you get that error, it indicates the program was unable to access the camera. In other words, the camera isn't working. The Pi Camera modules are tricky to set up, do you have a USB camera that you could try it with?
@PhatTran-cp8yf Ай бұрын
could you please Confirm ok or not with usb camera. I met the same problem on my own custom data
@BeraubtWerden Ай бұрын
Man when are you going to make how to deploy to android?
@Nj-hw2cv 23 күн бұрын
Use Google ML kit
@UngPKAYPREAK Ай бұрын
I step five, I got some error ValueError: Shape must be rank 3 but is rank 4 for '{{node train_input_images/write_summary/summary_cond/EncodePng}} = EncodePng[T=DT_UINT8, compression=-1](train_input_images/write_summary/summary_cond/strided_slice)' with input shapes: [3,320,320,3]. I'm sure how to solve this, anyone idea?
@UngPKAYPREAK Ай бұрын
I mean I'm sure sorry
@UngPKAYPREAK Ай бұрын
not sure