Finished improving my forest fire detection model. You can find all the information in the following medium post. Hope you enjoy it
cool project. I thought it was a good idea.
After finishing Lessons 1 and 2. I was looking for ideas for making a Classification model as a beginner.Found out that “Fox and Jackal” is one of the most confusing animal pairs in the world.
Why not use AI to do the hard thing for us? Let it be the animal expert for us.
So, built a Foxes vs Jackals Classifier. Check it out
- Next Step is to add Wolves, Hyenas etc to make it more interesting and see how the model performs.
Hi, everyone today i build a classifier using the code presented by @jeremy in Lesson 1, It’s called Is Young Adult or not.
I am interestingly facing another issue with my notebook after I wanted to rerun the whole cell and make some improvements.
My new problem is that when I decide to use vision_learner to train my model again, it gave me
NameError: name 'vision_learner' is not defined which is quite strange. I tried to import fastai.vision.all over again but I am still getting the same error.
The link is Spill classification | Kaggle
I am Harish from India.
I attended the first lecture and came up with a child vs elderly people classifier by modifying the code provided in the first lecture (to classify bird vs forest).
Here’s the kaggle notebook link: Is it a child? Creating a model from your own data | Kaggle
I could see lot of fun and interesting problems (I was jealous that I couldn’t think of one) people tried to solve using the first lecture. But I really found it difficult even to think about a problem. It is hard it seems to come up with a problem than coming up with a solution.
After two sessions of legend Jeremy Howard, I ended up with “Cat vs Dog classification using their voices”.
It’s my first project and I’m excited to share.Hope you guys love it.
I’ve competed in Kaggle for a while now but pretty happy with this result even though it’s not a top result…
The top public leaderboard results are around 0.801 and I’m at 0.78 with a pretty nominal training loop + using stratified k-fold cross-validation.
If you are giving Kaggle a try, don’t get discouraged. People make top notebooks public and if those notebooks happen to outperform your work, you will suddenly see 100s/1000s of people copying that notebook and making submissions with it to appear higher on the public leaderboard.
My internal score on held back test set
(this is evaluated using data held back from my training data)
I modified the Bird classification notebook created in Lesson 1 to make it generic.
Users can input Keywords (any) to train model and classify images on the go.
I have shared the notebook What is it? Multiclass Image Classification FastAI on Kaggle and also a Colab version What is it? Multiclass Image Classification FastAI.
Happy Learning !
After completing lesson 1 video and book chapter, I’ve made a berries classifier! You can check out the model’s performance to differentiate between cranberries and redcurrants here: lets classify berries ^_^ | Kaggle
Looking forward to learn more!
Well done solving your previous issue yourself. You learn a lot more searching on error messages rather than leaning on someone else to do it
I’ve sometimes experienced confusing results when the wrong version of fastai was loaded - ie I had fastai v1.x.x. loaded when trying a tutorial using a function introduced in v2.x.x. The confounding factor being that library versions are cached after they are loaded, so doing
!pip install -U fastai does nothing until a kernel restart. I don’t know the specifics of when
vision_learner was introduced, or if this relates at all to your issue, but its simple to try an upgrade followed by a kernel restart to see if it clears the error.
Also checking the version is sometimes a useful diagnostic
import fastai fastai.__version__
I’ve created the most useless model but it was fun to create it. i’ve modified is it a bird Model to is it a ‘anime person’ Model.
i don’t understand most of it and lot of time was spent on setting up Colab for this project and mounting google drive but i’ve made my first ever model:)
Source Code: Github Gist
So, I am new here and I just finished the first lecture. I have actually made a quite simple model which identifies Bears instead of Birds. I think somebody said it feels strange to run this much code without knowing what you’re doing. It got actually frustrating for me as I kept encountering the SameFileError but the forum at the end was very helpful.
Like I said, it is pretty simple and I would love to know how I can improve it.
Thank you @bencoman
Hey everyone. I recently completed the first lesson. For the homework, I decided to try and develop a weapon classifier. Crime is a big problem where I’m from and I’ve always thought that AI could be helpful to counteract it. I would love it if I pick up enough from this course to actually extend this and build a useful system.
The notebook can be found here.
I get a “404 not found” error. I guess you need to make your notebook public.
Sorry about that, I thought I did. Here it is.