YOLO Object Detection (TensorFlow tutorial)

YOLO Object Detection (TensorFlow tutorial)


Hello world, it’s Siraj. And YOLO, You Only Look Once.
That is the demo for Today’s Video.
We are gonna build Object Detection algorithm
Called YOLO, You Only Look Once.

100 thoughts on “YOLO Object Detection (TensorFlow tutorial)”

  1. Anyone know if a "Is this similar" YOLO type system has been built? As in, instead of training the CNN to classify certain objects you use a picture of the reference objects you are looking for, and the CNN is trained to identify if the reference object is in the scene or not? Would be much more useful for specific applications and not require millions of datapoints for each reference object, assuming the CNN has learnt to generalize on how to extract object similarity 🙂
    Want to build a object recognizer for my fridge, so it can keep track of what I have in my fridge, and I don't have to manually train it on every possible item I might put in the fridge, it will just keep track of different items I've placed in the fridge by using a reference image captured when I placed the object in there, then I can manually name the objects in my fridge (or just go by picture) when I want to know what I have in my fridge.

  2. Thanks for sharing, We also have some technical demo with this "OpenCV+ AI Yoloで飲食店でのクーポンアプリ" : https://onetech.jp/news-detail/OpenCV-AIYoloで飲食店でのクーポンアプリ/175

  3. @Siraj
    Hey.
    After running the ".flow –model cfg/yolo-ne

    w.cfg –load bin/yolo-new.weights –demo test.mp4" command, I just get a pop-up asking me to select how do I open the file.

  4. 😛 Three hours of hacking later it finally worked! It does not like smartphone video which has been recorded vertically (just flips it back) 😀

  5. Hi Siraj, How to show yolo object detection output on browser not from from pop-up window? I want to make it can be accessed from different IP.. I thought about using flask but I have no idea to do that.. really need guidance.. thanks

  6. Siraj Im getting this error

    'python3' is not recognized as an internal or external command,

    operable program or batch file.

    I'm a windows user, any help?

  7. Anyone got any opinions/warnings regarding YOLOv3? About to start a project and dont wanna make my life more difficult than it already is

  8. Hey, does anyone know what library was used for bounding box annotation? I think it looks very nice and I want to use it in my project but I can't seem to find it.

  9. Can I use YOLO but using only for text images detection? if so, where can I find it's dataset? Or should I train it from scratch?

  10. HOG is not really effective when it comes to different races (other than white people) – or else they could have included more features to suit most of the races

  11. Guys, do you know how privileged we are for having access to this? Thanks to Siraj, to the awesome people who made this possible and special thanks to capitalism!

  12. Hi Siraj. I've some doubt here. YOLO detects and recognizes it is a human or watch or cup etc. Why YOLO is called as detection not recognition?

  13. Please don't talk about the classical object detection algorithms if you do not understand them. Here is a list of some errors you made (not all of them):

    – Viola-Jones was using AdaBoost as a classifier and not SVM
    – Linear classifier and binary classifier are not the same thing. A linear classifier performs classification using a linear function, can be also multi class. A binary classifier just discriminates two classes, no multi class.
    – "They just downsample an image which in this case is actually a feature" – This is just so wrong. They were using Haar Features
    – You did not even mention the main contribution of the paper which is the cascade architecture.
    – HOG was not invented by Dalal and Triggs. They were just using the HOG feature descriptor in their object detection framework
    – Dalal and Triggs did not apply their algorithm on faces in their paper
    – You clearly do not understand the mathematical meaning of a gradient. They are using that word for this as it is the correct word. In both cases, HOG and backpropagation, gradients are computed.
    – But it is not about the gradients directly, it is about the orientation of gradients. And the histogram of these gradient orientation gives you the feature vector.
    – Dalal and Triggs did not use "some sort of similarity distance" to find new faces. In this case they actually used a linear Support Vector Machine for classification.
    – HOG is not very similar to what Convolutional Neural Networks learn themselves. Convolutional Kernerls are very different to HOG features. Gabor Filters are more similar. A convolutional neural network can not represent histograms of gradient orientations

    It is ok if you don't know these algorithms. But what it is very disappointing that you apparently don't understand the mathematical meaning of gradients being an AI / Deep Learning teacher.

  14. YOLO also outputs refined bounding box coordinates relative to the anchor boxes. And at the end non-maximum suppression is applied to avoid multi detections. You explain it as you would only threshold the confidence values and the anchor boxes would be the direct output.

  15. And you are also lying. From the console outputs you can see that you never got it running yourself. You are just showing a video.
    And you are also not training it on your own videos if you perform inference using pretrained weights.

  16. Help post: python: can't open file 'pip': [errno 2] no such file or directory . In Kali Linux . BTW, Siraj, You are the best! Mean it !

  17. Hello, I'm really interested in Computer vision … and recently engaged in python and trying GitHub code related to computer vision and its so fascinating. I have the idea in mind to combine object detection and barcode reader. What should I learn first to enable me to develop my future project? Your respond would really help my future career

  18. can i create a button a in tf classify and open a new empty activity .. co'z when i try it always crushed .. just asking if possible.. thank you!

  19. This is such an empty video. Is there a channel that actually talks about how these algorithms work? Instead of a ELI5 summary and implementing a github.

  20. Hi there, i am working in a project using YOLOv3, trying to detect whether workers on construction sites are wearing appropriate PPEs or not, you can have a look the progress at the moment here: https://youtu.be/LpUyQbC65EA

    I'll appreciate any comment.

    Regards.

  21. How can i get accuracy from YOLO model like any classifier? Is their is a possible way to do it?

  22. Thank you sir, I would try this, if it could run successfully, I will upload on Youtube in Indonesian Language

  23. HI sriraj .. i am looking for the model to extract phone numbers from the image . what method would be suitable for my need?

  24. I think it's developed by Joseph Redmon …… YOLO i've seen his TED talk. and also he made it as open source.

  25. I get an error:" utilsprocess.py", line 29, in parser

    if layer['type'] == '[net]':

    KeyError: 'type' " when loading the weights. I dont know whyyyy

  26. hi, in which part can you change the type of detect object? For example I would like to recognize a apple

  27. Cool thank you for the informative video. However, at the end you are not training it on your own video.. but you are only executing the neural network on your own video. Am I correct? You are only doing inference and not training…

  28. Hi guys can anyone help me? I installed the weights and pip install cpython ,but Im not sure how to run it

  29. Hi siraj,

    I want to know, is there a way to use object or image recognition on internet or website like on Amazon or Walmart to product images recognition, etc, rather than using real-time object detection via webcam. please reply

  30. I just wanted to say thank you for the great video I have been watching your videos for a little while now and I really enjoy the pacing of this one some of your other videos are a little fast and everything goes over my head but I think this was a good balance

  31. Great video @Siraj Raval !
    Is there a web version of the document you are going through in this video by any chance?

  32. Good video. awesome presentation. Tune your mike echo. it echo's at a considerable level, thereby muffling the sound clarity to a significant level. Use loud key statement phrases with a different tone, every so often, as headings for different elements of your great talk. Clear text uses paragraphs for different phases of a narrative. Your narrative is not loudly organized, it's quite fast and sonically uneventful. I am 40 and foreign, loud summary statements to orient and punctuate, and logical paragraphing, would make this clearer. awesome presentation though. I managed to follow at least 8 minutes until my potato sausages destracted me and when I picked up again i was lost.

Leave a Reply

Your email address will not be published. Required fields are marked *