295 - ONNX - open format for machine learning models
Ғылым және технология
ONNX - open format for machine learning models.
ONNX is a new standard for exchanging deep learning models.
It is an intermediary machine learning framework used to convert between different machine learning frameworks.
Useful blogs:
towardsdatascience.com/an-emp...
cloudblogs.microsoft.com/open...
Example scenario - You are deploying your trained model via an iOS app:
Modern Apple devices come with Apple Neural Engine (ANE) as part of the chip.
Deep learning models can work on CPU only, CPU+GPU, or all CPU/GPU/ANE
Fastest way is to run it on ANE, especially for heavy applications
The framework optimized for deep learning inference on iOS is CoreML
CoreML is built into the OS No need to compile, link, or ship binaries of ML libraries with the app.
Пікірлер: 33
Thank you very much Sir for your efforts. I am really grateful for your amazing guidance.
Great video! I am using ONNX on several projects and it has become part of my standard toolchain. It is amazing how just a few years ago even installing cuDNN was a chore and now we have a universal instruction set for ML models.
@DigitalSreeni
Жыл бұрын
Thanks again Yujan. Using deep learning for various applications has become very easy, even for non-experts. I am glad Facebook, Google and Microsoft are open sourcing their code.
I am really grateful, you chose this topic. I have one request , if you can make tutorial on using ONNX with cuda, It will be extremely helpful. Thanks !
I would love to see more on CoreML. In any case, as always great videos! you are a master communicator!
Tks a looot for this upcoming series of videos. They´re right on time for me. I actually need to use ONNX at my job in the next days. Looking forward to your next videos Mr. Sreeni. tks again for sharing your knowledge
A multiclass semantic segmentation video with knowledge distillation training (Teacher-Student Network) would be very nice.
Thank you
Great video
Please have you coverd before graph convolutional network?
Hey Sreeni, I would love to see more on object tracking between different frames by using some advanced techniques with object detection. I tried tracy but it is not that effective.
¡Gracias!
@DigitalSreeni
Жыл бұрын
Muchas gracias mi amigo!
Multivariate time-series using transformers would be nice complement to your multivariate TS using LSTM video.
awesome video!! learned a lot from here. btw Could you make video for autoencoder? A practical video on where and how to use it.
hi sir！ recently, i habe trying to convert some older trained models to the onnx format. but i have been getting different errors during different model conversions. for example: some model trained using pytorch ==0.6.0 or keras==2.1.5 (few year older versions). i am concluding that older version trained models cannot be converted to onnx, forcing me to train again now with newer version models. if any one else experienced this behaviour. do reply in this comment plz.
Great Work, please make a video on Lung cancer image segmentation using LIDC-IDRI dataset (dicom images). Thanks
I would love to see Semantic segmentation of satellite imagery using U-net using Ensemble of networks (e.g., unet and Res-net ) - multiclass and binary
Hi Digital Sreeni. Can you maje a video on advanced image registration using DL such as Voxelmorph. Thank you
Forget iOS...can you show how onnx flow works on an S24?
Hi Digital Sreeni. How about an Android Application? can you make a video in advance on both (IOS and Android) Thank you for this video, I'm looking forward to the ONNX series of videos.
Hi Sreeni, I would like to see you talk about JAX if you can
Great
Hello, do you have a video that cover topic of generating pyramidal images? If not can you make one? I'm having huge trouble with this.
Hi Sreeni, is it possible to make tutorials on SlideFlow package?
Digital signal processing can you post some videos about it
Hi sir ... Can you please upload a video about how to use metahuristic optimization algorithms for IMAGE CLASSIFICATION especially for medical imaging?
@DigitalSreeni
Жыл бұрын
I am not familiar with that field, especially for image classification. All I know is that metahuristic algorithms can be used for optimization purposes, for example, tuning hyperparameters.
@vidyasvidhyalaya
Жыл бұрын
@@DigitalSreeni Okay sir. Then please upload a video about hypertune the parameters by using image datasets sir 🙏
Can you please make a vid how yo initialize .onnx
Can you make some videos on TinyML/Embedded Machine Learning?
Hello, I really appreciate your nice and efficient work and I really leant a lot from you videos thanks. I really need a Video on GI track segmentation which is the Kaggle competition. thanks in advance. please let me know how to donate in comment section.
sir can i get ur email id , i have certain doubts in segmentation of aerial images videos it would be really helpful