The homeworks can easily be cloned from one of several github repositories after a quick google search + the certificate may seem fancy but industry doesn't seem to care much about these types of certificates when hiring
Bob Crunch, you can Audit this course from the Coursera website, which basically means you can watch the videos for free but don't get access to the quizzes and assignments.
@@sourabhkhandelwal689 I have audited the Coursera class. I've copied the code from the course Jupyter Notebooks and have screenshots of the other information on the Notebooks. The only thing missing is the datasets. You can search the 'Net for equivalent datasets; e.g., I found a hand sign language dataset (sign-language-digits-dataset) that's equivalent to the courses dataset. I found a converter that converts the numerical arrays to an JPG (package imageio). It will take an effort to adapt the code, but it's a good learning experience.
Yes you can. In pytorch for the faster r-cnn framework with torchvision you can input a shape of any size, and the GeneralTransform module will rescale it to a fixed min and max size. Convolution operations do not require a certain spatial size as input, although you will see in Tensorflow that they often have checks for the input shape size that do restrict the shape of the image, however this is not necessary in my opinion. The ROI pool operation from the Faster R-CNN architecture gives you a fixed size tensor no matter what your input shape is (assuming it has 3 channels as in an RGB image of course).
![The moment we stopped understanding AI [AlexNet]](http://i.ytimg.com/vi/UZDiGooFs54/mqdefault.jpg)
![Gucci Mane & Sexyy Red - You Don't Love Me [Official Music Video]](http://i.ytimg.com/vi/j7IBqyFKT4c/mqdefault.jpg)
The woman in the picture for "neural style transfer" is actually his wife. :)
But he is not going to transfer her to you in anycase !
I feel so lucky to find out these videos, thank you.
Anybody starting, I suggest you use 1.5x Playback speed. It should save your time a lot. Really.
Are amphetamines now legal in Israel? Because it's almost unintelligable at 1.5x :-D :-D
Already watching at 2.3x... 😂😂
@@MSDOS128 Good job, I'm using 500X speed, it really saved a lot of time
nice advice
My best teacher
playlist is 6hour, 1minute, 12second.
Is the volume too low?
This is a very informative course. Can I get the slides?
just screenshot them!
You can get them for free from the official website.
@@saiefzneti from where i can get slides
sen nasıl bir kralsın ya, youtubea yüklemiş videoları
说的些啥了
pata ni yo ke bolny aen tou
Pora picchi pulka
for some reason there were Korean subtitles on for default.
RUclips is racist
Korea's passion about AI
Why pay the exact same content on Coursera?? I do!
The homeworks can easily be cloned from one of several github repositories after a quick google search + the certificate may seem fancy but industry doesn't seem to care much about these types of certificates when hiring
certificate
I'm taking the class right now, and some of the course videos are missing from this playlist, e.g. the videos on objects and bounding boxes.
Bob Crunch, you can Audit this course from the Coursera website, which basically means you can watch the videos for free but don't get access to the quizzes and assignments.
@@sourabhkhandelwal689 I have audited the Coursera class. I've copied the code from the course Jupyter Notebooks and have screenshots of the other information on the Notebooks. The only thing missing is the datasets. You can search the 'Net for equivalent datasets; e.g., I found a hand sign language dataset (sign-language-digits-dataset) that's equivalent to the courses dataset. I found a converter that converts the numerical arrays to an JPG (package imageio). It will take an effort to adapt the code, but it's a good learning experience.
can we use 2MB image to object detection and image classification if model is trained for 256by256 pixel images
In convention, CNN model's input has to reshape into the specific scale before start computation.
i would lower the resolution first
Yes you can. In pytorch for the faster r-cnn framework with torchvision you can input a shape of any size, and the GeneralTransform module will rescale it to a fixed min and max size. Convolution operations do not require a certain spatial size as input, although you will see in Tensorflow that they often have checks for the input shape size that do restrict the shape of the image, however this is not necessary in my opinion. The ROI pool operation from the Faster R-CNN architecture gives you a fixed size tensor no matter what your input shape is (assuming it has 3 channels as in an RGB image of course).
niceexplanation
The best
it is a good video,, ;)
sensei
i need the programming exercices for specialisation courses coursera (zip file)
Search on google, "Deep learning andrew ng github" and you'll get maany github repos
@@ojasrox but what I need is the exercises (zip files) not the solution